Error in running asteroid\egs\musdb18\X-UMX\train.py

## 🐛 Bug

Due to the fact that the 1.5.0 version of pytorch-lightning removed the distributed backend, some erorrs occured when running `asteroid\egs\musdb18\X-UMX\train.py`

### To Reproduce

`cd asteroid\egs\musdb18\X-UMX`
`python train.py`

### Expected behavior
Error message:

`Traceback (most recent call last): File "train.py", line 500, in main(arg_dic, plain_args) File "train.py", line 459, in main trainer = pl.Trainer( File "C:\Users\Huicheng Zhang\.conda\envs\xumx\lib\site-packages\pytorch_lightning\trainer\connectors\env_vars_connector.py", line 38, in insert_env_defaults return fn(self, **kwargs) TypeError: __init__() got an unexpected keyword argument 'distributed_backend'`

### Environment

#### Package versions

Run `asteroid-versions` and paste the output here:

Asteroid 0.5.1
PyTorch 1.10.0+cpu
PyTorch-Lightning 1.5.0

#### Additional info

After reinstalling pytorch-lightning to 1.4.9 the above problem will be solved. However it would be good to upgrade the code to PL 1.5.0

Start with asteroid/egs/musdb18/X-UMX/train.py, especially the Trainer setup around line 459, and reproduce the issue with `python train.py` using PyTorch-Lightning 1.5.0. Done means the training entry point no longer passes the removed distributed backend argument and starts without the reported TypeError.

Written by the indexing model from the issue text.