Add dashboard df export

[sfmig]

Added dataframe export functionality to the dashboard tab.

The user can export a subset of the available pose data by:

• selecting a subset of videos (for each of them there exists a corresponding h5 file with the pose data), and
• selecting a subset of frames per video, based on the event tags.
  A dataframe is exported combining these two requirements. I added mock event tags to the sample project to check it works as intended.

Features:

• If the pose data is not available for a video, the row will be automatically unselected and a message will be shown.
• some extra columns are added to the exported dataframe, apart from those provided by DLC. These are 'event_tag' and 'ROI'.
  • 'Event_tag' assigns the 'event_tag' label to all frames from the instant in which the event is defined until a new event happens.
  • For now, the ROI label is just an empty string, but the idea would be to read the ROI path from the video metadata file and check if a bodypart at a specific frame is inside the ROI.
• I added the 'Events' fields to the yaml file describing the metadata fields (otherwise the tests failed, thanks @samcunliffe 😇 ). To make it compatible with the table visualisation in the metadata tab, I modified it to show any field that is a dict ('Events' and 'ROI') as a string. I think for now it's ok, but we should make these columns not editable in the metadata tab.

Caveats

• I'm exporting the dataframe using h5 files for now, but happy to consider other options going forward (mainly thinking about the corruption issues @niksirbi and @adamltyson mentioned a while ago).
• it is a bit ambiguous to use the term 'event_tag' for an instant (i.e. a specific frame) and a period (in the exported dataframe). Should I use different terms?
• the 'event_tag' in the exported dataframe is added for each bodypart. This is repetitive, since the event_tags refer only to the frame number and don't care about bodyparts, but I was struggling to get this done 'properly' with pandas. If anyone has pandas expertise I'm all ears, it has been driving me mad 😅 🐼 👎

Pending bits:

• issue add offset from the event tags #39
• issue add ROI in the exported dataframe #40 (right now it's just a place holder)
• add plots:
  • trajectory plots, with different colour options (colour by video, by frame, by bodypart, etc)
  • heatmap plots
• in metadata tab: make columns for 'Events' and 'ROI' not editable --issue fixes to metadata tab #41

[codecov-commenter]

Codecov Report

Merging #38 (b45cd8f) into main (9ae154c) will decrease coverage by 6.37%.
The diff coverage is 30.24%.

@@            Coverage Diff             @@
##             main      #38      +/-   ##
==========================================
- Coverage   40.39%   34.03%   -6.37%     
==========================================
  Files           9       11       +2     
  Lines         458      526      +68     
==========================================
- Hits          185      179       -6     
- Misses        273      347      +74     

Impacted Files	Coverage Δ
wazp/pages/home.py	100.00% <ø> (ø)
wazp/utils.py	28.31% <8.69%> (-4.68%)	⬇️
wazp/callbacks/metadata.py	25.39% <25.39%> (ø)
wazp/callbacks/home.py	26.66% <26.66%> (ø)
wazp/callbacks/dashboard.py	28.37% <28.37%> (ø)
wazp/app.py	95.00% <100.00%> (+0.88%)	⬆️
wazp/callbacks/roi.py	16.86% <100.00%> (ø)
wazp/pages/04_dashboard.py	100.00% <100.00%> (ø)

... and 1 file with indirect coverage changes

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[niksirbi]

I found nothing critical.
I noticed that a fair bit of refactoring has crept into this PR (like dividing callbacks into separate files). This was sorely needed, but it may make it hard for you to merge into main now.

You may either choose to power through and resolve conflicts or try to transfer the refactoring bits to a separate PR. Whatever is easiest for you.

I would also remove the bits about the custom plots, since we are not going down that road after all.

[samcunliffe]

So it all looks good and the pandas.df export works 👍 .

My substantial comments are:

1. if we really want to remove the auto plots from the dashboard (if we do this, we're not really a dashboard any more) and..
2. can we have the suggested script download as a .py (or something) rather than a textbox.

[sfmig]

Implemented the smaller fixes, and answered the larger requests with comments. I will take care of the bigger bits on separate PRs (following our short PR aims), but do let me know if you'd prefer otherwise.

[samcunliffe]

Conflicts!