Can't cache sampler_nowcaster_test

### Context
After successfully running
`python forecast_demo.py` and
`python train_autoenc.py --model_dir="../models/autoenc_train"`,
I couldn't get the
`python train_genforecast.py --model_dir="../models/genforecast_train"`
command running due to problems in caching the sampler for test and training set.

When running `python train_genforecast.py --model_dir="../models/genforecast_train"`:
### Expected behaviour
1. Creates the sampler files and save to the `cache` directory for valid, test, and train datasets
2. Trains the forecaster

### Actual behaviour
1. Creates the file `../cache/sampler_nowcaster_valid.pkl`
2. Throws an error creating the next sampler as (complete error message pasted below):
```
~/tmp/0606/ldcast/scripts$ python train_genforecast.py --model_dir="../models/genforecast_train"
Loading data...
/home/kucuk/tmp/0606/ldcast/ldcast/features/transform.py:80: RuntimeWarning: divide by zero encountered in log10
log_scale = np.log10(scale).astype(np.float32)
Loading cached sampler from ../cache/sampler_nowcaster_valid.pkl.
No cached sampler found, creating a new one...
Traceback (most recent call last):
File "train_genforecast.py", line 129, in
Fire(main)
File "/home/kucuk/miniconda3/envs/ldcast_test/lib/python3.8/site-packages/fire/core.py", line 141, in Fire
component_trace = _Fire(component, args, parsed_flag_args, context, name)
File "/home/kucuk/miniconda3/envs/ldcast_test/lib/python3.8/site-packages/fire/core.py", line 475, in _Fire
component, remaining_args = _CallAndUpdateTrace(
File "/home/kucuk/miniconda3/envs/ldcast_test/lib/python3.8/site-packages/fire/core.py", line 691, in _CallAndUpdateTrace
component = fn(*varargs, **kwargs)
File "train_genforecast.py", line 125, in main
train(**config)
File "train_genforecast.py", line 94, in train
datamodule = setup_data(
File "/home/kucuk/tmp/0606/ldcast/scripts/train_nowcaster.py", line 124, in setup_data
datamodule = split.DataModule(
File "/home/kucuk/tmp/0606/ldcast/ldcast/features/split.py", line 127, in __init__
self.batch_gen = {
File "/home/kucuk/tmp/0606/ldcast/ldcast/features/split.py", line 128, in
split: batch.BatchGenerator(
File "/home/kucuk/tmp/0606/ldcast/ldcast/features/batch.py", line 81, in __init__
self.sampler = EqualFrequencySampler(
File "/home/kucuk/tmp/0606/ldcast/ldcast/features/sampling.py", line 30, in __init__
self.starting_ind = [
File "/home/kucuk/tmp/0606/ldcast/ldcast/features/sampling.py", line 31, in
starting_indices_for_centers(
File "/home/kucuk/tmp/0606/ldcast/ldcast/features/sampling.py", line 210, in starting_indices_for_centers
starting_ind = np.concatenate(
File "<__array_function__ internals>", line 180, in concatenate
ValueError: need at least one array to concatenate
```
3. Seems like an issue with indexing of patches in the sampler, though I'm not sure...

### Additional information

- I removed the `sampler_nowcaster*` in the `cache` folder and tried the same command `python train_genforecast.py --model_dir="../models/genforecast_train"`, received the same error
- In case it helps, below is the complete directory structure (note the size difference among `sampler_*_valid.pkl` files)
```
~/tmp/0606/ldcast$ tree -h
.
├── [ 11K] LICENSE
├── [5.5K] README.md
├── [4.0K] cache
│   ├── [501K] sampler_autoenc_test.pkl
│   ├── [883M] sampler_autoenc_train.pkl
│   ├── [1.9M] sampler_autoenc_valid.pkl
│   └── [745K] sampler_nowcaster_valid.pkl
├── [4.0K] config
│   ├── [ 65] genforecast-radaronly-128x128-20step.yaml
│   └── [ 158] genforecast-radaronly-256x256-20step.yaml
├── [4.0K] data
│   ├── [ 48K] Border_CH.dbf
│   ├── [130K] Border_CH.shp
│   ├── [4.0K] RV
│   │   ├── [197M] patches_RV_202204.nc
│   │   ├── [156M] patches_RV_202205.nc
│   │   ├── [154M] patches_RV_202206.nc
│   │   ├── [151M] patches_RV_202207.nc
│   │   ├── [ 84M] patches_RV_202208.nc
│   │   └── [281M] patches_RV_202209.nc
│   ├── [4.0K] RZC
│   │   ├── [ 55M] patches_RZC_201804.nc
│   │   ├── [102M] patches_RZC_201805.nc
│   │   ├── [ 53M] patches_RZC_201806.nc
│   │   ├── [ 53M] patches_RZC_201807.nc
│   │   ├── [ 71M] patches_RZC_201808.nc
│   │   ├── [ 38M] patches_RZC_201809.nc
│   │   ├── [ 72M] patches_RZC_201904.nc
│   │   ├── [117M] patches_RZC_201905.nc
│   │   ├── [ 69M] patches_RZC_201906.nc
│   │   ├── [ 57M] patches_RZC_201907.nc
│   │   ├── [ 77M] patches_RZC_201908.nc
│   │   ├── [ 55M] patches_RZC_201909.nc
│   │   ├── [ 42M] patches_RZC_202004.nc
│   │   ├── [ 67M] patches_RZC_202005.nc
│   │   ├── [110M] patches_RZC_202006.nc
│   │   ├── [ 51M] patches_RZC_202007.nc
│   │   ├── [ 91M] patches_RZC_202008.nc
│   │   ├── [ 61M] patches_RZC_202009.nc
│   │   ├── [ 59M] patches_RZC_202104.nc
│   │   ├── [140M] patches_RZC_202105.nc
│   │   ├── [ 86M] patches_RZC_202106.nc
│   │   ├── [120M] patches_RZC_202107.nc
│   │   ├── [ 62M] patches_RZC_202108.nc
│   │   └── [ 54M] patches_RZC_202109.nc
│   ├── [4.0K] demo
│   │   └── [4.0K] 20210622
│   │   ├── [214K] RZC211731820VL.801.h5
│   │   ├── [214K] RZC211731825VL.801.h5
│   │   ├── [216K] RZC211731830VL.801.h5
│   │   └── [216K] RZC211731835VL.801.h5
│   └── [4.1K] split_chunks.pkl.gz
├── [4.0K] figures
│   └── [4.0K] demo
│   ├── [207K] R_past-00.png
│   ├── [208K] R_past-01.png
│   ├── [208K] R_past-02.png
│   ├── [208K] R_past-03.png
│   ├── [192K] R_pred-00.png
│   ├── [191K] R_pred-01.png
│   ├── [192K] R_pred-02.png
│   ├── [194K] R_pred-03.png
│   ├── [195K] R_pred-04.png
│   ├── [195K] R_pred-05.png
│   ├── [196K] R_pred-06.png
│   ├── [196K] R_pred-07.png
│   ├── [199K] R_pred-08.png
│   ├── [202K] R_pred-09.png
│   ├── [204K] R_pred-10.png
│   ├── [204K] R_pred-11.png
│   ├── [208K] R_pred-12.png
│   ├── [207K] R_pred-13.png
│   ├── [211K] R_pred-14.png
│   ├── [211K] R_pred-15.png
│   ├── [212K] R_pred-16.png
│   ├── [209K] R_pred-17.png
│   ├── [209K] R_pred-18.png
│   └── [205K] R_pred-19.png
├── [4.0K] ldcast
│   ├── [4.0K] analysis
│   │   ├── [4.8K] crps.py
│   │   ├── [4.4K] fss.py
│   │   ├── [3.2K] histogram.py
│   │   └── [5.5K] rank.py
│   ├── [4.0K] features
│   │   ├── [4.0K] __pycache__
│   │   │   ├── [ 11K] batch.cpython-38.pyc
│   │   │   ├── [ 11K] patches.cpython-38.pyc
│   │   │   ├── [7.0K] sampling.cpython-38.pyc
│   │   │   ├── [4.8K] split.cpython-38.pyc
│   │   │   ├── [8.2K] transform.cpython-38.pyc
│   │   │   └── [3.1K] utils.cpython-38.pyc
│   │   ├── [ 13K] batch.py
│   │   ├── [3.9K] io.py
│   │   ├── [ 13K] patches.py
│   │   ├── [7.1K] sampling.py
│   │   ├── [5.2K] split.py
│   │   ├── [8.7K] transform.py
│   │   └── [3.8K] utils.py
│   ├── [8.5K] forecast.py
│   ├── [4.0K] models
│   │   ├── [4.0K] __pycache__
│   │   │   ├── [1.1K] distributions.cpython-38.pyc
│   │   │   └── [ 833] utils.cpython-38.pyc
│   │   ├── [4.0K] autoenc
│   │   │   ├── [4.0K] __pycache__
│   │   │   │   ├── [3.4K] autoenc.cpython-38.pyc
│   │   │   │   ├── [1.9K] encoder.cpython-38.pyc
│   │   │   │   └── [ 960] training.cpython-38.pyc
│   │   │   ├── [3.0K] autoenc.py
│   │   │   ├── [1.9K] encoder.py
│   │   │   └── [ 952] training.py
│   │   ├── [4.0K] benchmarks
│   │   │   ├── [2.4K] dgmr.py
│   │   │   ├── [3.7K] pysteps.py
│   │   │   └── [ 350] transform.py
│   │   ├── [4.0K] blocks
│   │   │   ├── [4.0K] __pycache__
│   │   │   │   ├── [9.5K] afno.cpython-38.pyc
│   │   │   │   ├── [3.2K] attention.cpython-38.pyc
│   │   │   │   └── [2.2K] resnet.cpython-38.pyc
│   │   │   ├── [ 13K] afno.py
│   │   │   ├── [3.0K] attention.py
│   │   │   └── [2.7K] resnet.py
│   │   ├── [4.0K] diffusion
│   │   │   ├── [4.0K] __pycache__
│   │   │   │   ├── [6.6K] diffusion.cpython-38.pyc
│   │   │   │   ├── [2.9K] ema.cpython-38.pyc
│   │   │   │   └── [8.3K] utils.cpython-38.pyc
│   │   │   ├── [7.4K] diffusion.py
│   │   │   ├── [2.9K] ema.py
│   │   │   ├── [ 12K] plms.py
│   │   │   └── [8.7K] utils.py
│   │   ├── [ 838] distributions.py
│   │   ├── [4.0K] genforecast
│   │   │   ├── [4.0K] __pycache__
│   │   │   │   ├── [1.4K] analysis.cpython-38.pyc
│   │   │   │   ├── [1.0K] training.cpython-38.pyc
│   │   │   │   └── [ 11K] unet.cpython-38.pyc
│   │   │   ├── [1.0K] analysis.py
│   │   │   ├── [1.1K] training.py
│   │   │   └── [ 17K] unet.py
│   │   ├── [4.0K] nowcast
│   │   │   ├── [4.0K] __pycache__
│   │   │   │   └── [8.4K] nowcast.cpython-38.pyc
│   │   │   └── [8.3K] nowcast.py
│   │   └── [ 770] utils.py
│   └── [4.0K] visualization
│   ├── [1.2K] cm.py
│   └── [ 11K] plots.py
├── [4.0K] ldcast.egg-info
│   ├── [5.9K] PKG-INFO
│   ├── [ 175] SOURCES.txt
│   ├── [ 1] dependency_links.txt
│   ├── [ 137] requires.txt
│   └── [ 1] top_level.txt
├── [4.0K] models
│   ├── [4.0K] autoenc
│   │   └── [1.5M] autoenc-32-0.01.pt
│   ├── [4.0K] autoenc_train
│   │   ├── [4.6M] epoch=0-val_rec_loss=0.2204.ckpt
│   │   ├── [4.6M] epoch=0-val_rec_loss=nan.ckpt
│   │   └── [4.6M] epoch=1-val_rec_loss=0.1653.ckpt
│   └── [4.0K] genforecast
│   └── [5.0G] genforecast-radaronly-256x256-20step.pt
├── [4.0K] results
├── [4.0K] scripts
│   ├── [4.0K] __pycache__
│   │   └── [4.0K] train_nowcaster.cpython-38.pyc
│   ├── [4.0K] dwd_dataset.py
│   ├── [1.4K] eval_data.py
│   ├── [1.5K] eval_dgmr.py
│   ├── [4.4K] eval_genforecast.py
│   ├── [1.4K] eval_pysteps.py
│   ├── [3.8K] forecast_demo.py
│   ├── [4.0K] lightning_logs
│   │   └── [4.0K] version_0
│   │   ├── [4.0K] events.out.tfevents.1686053670.{VM_NAME}
│   │   └── [ 3] hparams.yaml
│   ├── [3.5K] metrics.py
│   ├── [ 13K] plots_genforecast.py
│   ├── [2.8K] train_autoenc.py
│   ├── [3.4K] train_genforecast.py
│   └── [4.1K] train_nowcaster.py
├── [ 951] setup.py
└── [ 0] tmp4_0607

37 directories, 149 files

```

Could it be the sth related to version compatibilities of packages, e.g., dask or numba? Perhaps I'm missing something in the `data` directory.
@jleinonen please let me know how I can provide further information - and thanks in advance!

Reproduce the failure with scripts/train_genforecast.py and follow the stack through scripts/train_nowcaster.py, ldcast/features/split.py, batch.py, and sampling.py. Start at starting_indices_for_centers in ldcast/features/sampling.py and inspect why concatenation receives no arrays for the test or train split. Done means the command creates valid, test, and train sampler caches and proceeds to forecaster training.

Written by the indexing model from the issue text.