RFC: get_channel_types polysemy

I’ve stumbled on a not-so-nice aspect of the API when working on #7486:

1. mne.io.pick.get_channel_types (source) takes no input, gets all known channel types as dict keys (values are FIFF constants for kind and coil_type).
2. mne.io.pick._get_channel_types (source) takes an Info, picks, boolean unique, and boolean whether to restrict to data channels only. Returns a list or set depending on unique.
3. <Raw/Epochs/Evoked>.get_channel_types (source) method (via the ContainsMixin). Returns list of channel types.

Number 1 is called once in a test, and nowhere else in the codebase. I’d like to propose renaming it from get_channel_types to get_all_channel_types, show_all_channel_types, or get_channel_types_dict. I prefer the first option. This is a public function so it would be an API change w/ deprecation cycle. If acceptable, I could roll it into #7486 or do it in a separate PR.

[drammock]

Also, I haven’t checked the return values of mne.io.pick.get_channel_types and mne.io.pick._get_channel_types, but can/should we de-duplicate code by calling the private function inside the public one?

Not possible, they do completely different things, which is why I proposed to rename one of them.

What I have done is deduplicate the instance methods and the private function, which were doing (roughly) the same thing.

[agramfort]

this is now what the PR from Dan does. Except using a short function name