Channel adjacency improvements

Proposed documentation enhancement

The docstring of the picks argument in read_ch_adjacency is wrong as there is no info provided to this function.

Also, find_ch_adjacency could benefit from an extra argument similar to pick in read_ch_adjacency to select channels to include in the adjacency matrix. I see 2 options forward:

• (1) Keep ch_type and add exclude=() which can be set to exclude="bads"
• (2) Deprecate ch_type in favor of the traditional picks argument, with an adapted docstring mentioning that the channel selection provided must yield a single channel type (+ check and appropriate error message)

I'm in favor of (2) as it offers more flexibility to find_ch_adjacency.

Correspond to this forum post: https://mne.discourse.group/t/find-ch-adjacency-not-accounting-for-bad-channels/7968/

Start with the read_ch_adjacency and find_ch_adjacency entry points and their linked documentation, then review the related forum discussion. Resolve the proposed channel-selection API before changing the picks docstring; done means the documentation is accurate and find_ch_adjacency supports the agreed selection behavior.