Fix channel order bug and make select_channels private

[chrishalcrow]

Fixes a channel order bug in the sparsity mask for the new select_channels method for analyzers.

Discussed with @alejoe91 and decided to keep this method private for now - we hope users will access through the split_by method rather than directly.

Fixed waveforms and pca selection.

[alejoe91]

Renaming

[alejoe91]

Another couple of minor comments :)

[alejoe91]

Suggested change

	channel_indices = []
	unit_channel_ids = old_unit_id_to_channel_ids[unit_id]
	for channel_id in channel_ids:
	if channel_id in unit_channel_ids:
	idx = np.where(old_unit_id_to_channel_ids[unit_id] == channel_id)[0][0]
	channel_indices.append(idx)
	unit_channel_ids = old_unit_id_to_channel_ids[unit_id]
	channel_ids_in_unit = [ch for ch in channel_ids if channel_id in unit_channel_ids]
	channel_indices = self.sorting_analyzer.channel_ids_to_indices(channel_ids_in_unit)

I think that this does the exact same thing, right?

[chrishalcrow]

You need the old channel_ids_to_indices, not the new ones from self.sorting_analyzer.channel_ids_to_indices

[alejoe91]

self is the old analyzer, no?

See:

old_waveforms = self.data["waveforms"]

[chrishalcrow]

Oop - we're both wrong. The channel_indices tell you how to map between the channel_indices with respect to the sparsity mask for each unit (no the overall channel_indices). You can make the code look very simple if you have access to the new sorting_analyzer's sparsity mask, which we don't easily have.

I've updated the variables names, which I think should help avoid confusion.

[alejoe91]

Fixes #4700

[alejoe91]

Perfetto! Just renamed a couple of variables in the _select_channels_sparse_data function.

The added tests are bullet-proof!