Preprocessing capsule fails with TypeError when zarr NWB stores starting_time as a 1-d array

Hi,
I'm getting an error with the preprocessing capsule when saving preprocessed data, if the input NWB file was written by `hdmf_zarr` with `starting_time` stored as a shape-(1,) array rather than a true scalar. The failure occurs deep in `SpikeInterface's BinaryRecordingExtractor.__init__`:

> "TypeError: only 0-dimensional arrays can be converted to Python scalars"

### Technical Context:
- Pipeline: aind-ephys-pipeline (revision `07b42a1a04`)
- Capsule: aind-ephys-preprocessing (commit 6a093b1)
- Container: `ghcr.io/allenneuraldynamics/aind-ephys-preprocessing:si-0.103.2`
- SpikeInterface version: 0.103.2
- Input format: Zarr-backed NWB (.nwb read via SI's zarr NWB backend)

### Issue Details
The zarr NWB backend reads starting_time from the zarr store using:
```
# spikeinterface/extractors/nwbextractors.py
t_start = electrical_series["starting_time"][()]
```
When hdmf_zarr writes an NWB file, it marks scalar fields with a zarr_dtype: scalar attribute, but older versions of the library (present in the si-0.103.2 container) do not honor this attribute on read. As a result, [()] returns a numpy ndarray of shape (1,) instead of a 0-d array or Python float.

This propagates to recording_processed.save(), which internally constructs a BinaryRecordingExtractor and calls:
```
# spikeinterface/core/binaryrecordingextractor.py, line 84
t_starts = [float(t_start) for t_start in t_starts]
```
Python's built-in float() raises a TypeError for any array with ndim >= 1.

Full traceback:
```
File "capsule/code/run_capsule.py", line 469, in
recording_bin = recording_processed.save(folder=preprocessing_output_folder)
File ".../spikeinterface/core/base.py", line 865, in save
loaded_extractor = self.save_to_folder(**kwargs)
File ".../spikeinterface/core/base.py", line 972, in save_to_folder
cached = self._save(folder=folder, verbose=verbose, **save_kwargs)
File ".../spikeinterface/core/baserecording.py", line 595, in _save
binary_rec = BinaryRecordingExtractor(
File ".../spikeinterface/core/binaryrecordingextractor.py", line 84, in __init__
t_starts = [float(t_start) for t_start in t_starts]
File ".../spikeinterface/core/binaryrecordingextractor.py", line 84, in
t_starts = [float(t_start) for t_start in t_starts]
TypeError: only 0-dimensional arrays can be converted to Python scalars
```