MAINT: Split bad-finding step

[larsoner]

Before merging …

• Separate bad-finding into its own step
• Add raw data plots to report (before preproc)
• Fix logic with exec_params by including it in the kwargs, but excluding it from cache via ignore arg in memory.cache(...).
• Restore grand-average source space report that was missing/dropped accidentally
• Add bad-channel finding and use for empty-room data
• Changelog has been updated (docs/source/changes.md)

Closes #657

It (hopefully) simplifies a bit the file logic. Let's see what the CIs say.

[larsoner]

Okay @hoechenberger this one is finally green and ready to go I think.

Some stuff might be simplified by having _01_data_quality.py write out a new _raw.fif with the break annotations, bads, etc. applied to it. But I wasn't sure if the extra storage space required would be worth it. Maybe in some follow-up PR...

[hoechenberger]

Some stuff might be simplified by having _01_data_quality.py write out a new _raw.fif with the break annotations, bads, etc. applied to it. But I wasn't sure if the extra storage space required would be worth it.

We could use write_raw_bids() with symlink=True but yes, that's probably something for a followup PR!

[hoechenberger]

Thanks @larsoner!