Fix report section on original raw data (didn't contain bad channels and subject or experimenter name)

[hoechenberger]

Fixes #930

Before merging …

• Changelog has been updated (docs/source/changes.md)

[hoechenberger]

@larsoner This seems to fix my problem

I only have one question

Should we only set the pre-existing bads when generating the HTML repr for the original data, or should we set the pre-existing bads + all bads from the data quality checks?

Either approach seems to have its up- and downsides.

cc @SophieHerbst

[hoechenberger]

This change should avoid the problem we saw here, as it now forces us to provide pre-loaded raw data (with hopefully the correct metadata set). previously this was optional. see also the changes below

[hoechenberger]

Circle fails for ds000246 due to outdated cache

[larsoner]

I'll fix the caches in #929

[hoechenberger]

I'll fix the caches in #929

Thank you!

[larsoner]

Should we only set the pre-existing bads when generating the HTML repr for the original data, or should we set the pre-existing bads + all bads from the data quality checks?

To me the optimal thing to do here I think is to have two report entries:

1. One which is the data as it comes originally from the dataset author with bads from the sidecar. It tells you what you start with before doing anything to it with the pipeline
2. Another which is the data after our autobads (this section only needs to exist if we do some autobad detection). It tells you what the automated approaches did to the data

To me it's okay for brevity to have (1) have all the nice raw data plots, and have (2) maybe have less -- like just a list of channels removed or something? I think we're already kind of close to this organization:

But from that data quality plot it's not immediately clear which channels were added as bad, if any. Even just a HTML entry that lists the bads (when bad enabled) and flats (when flat enabled) would be a nice help

[larsoner]

... and organizationally I think it might make sense to call that section in the sidebar Raw (autobad) or similar. It makes it clear the parentheticals from original->autobad->maxwell->filtered->cleaned. "Data Quality" is an outlier at the moment in that it uses the step name

[hoechenberger]

Thanks for sharing your ideas, they resonate well with me!

I believe then at least the bits of code that this PR touches should be good! And the rest could be addressed in future PRs

I will add a changelog entry and once CI passes, we should be good to go.

[larsoner]

I merged main into the branch, you should be able to add to the changelog now and it should come back green

[hoechenberger]

Thanks @larsoner, I pushed changelog entries and enabled auto-merge