Add fail-fast for datasets outside the visible extent

[raphaelquast]

While trying to improve datashader integration with EOmaps I ran into the following problem:

At the moment, DSArtist.make_image() does not check if the data is actually contained within the visible extent before attempting to create an image.

This results in a huge performance penalty if a large number of datasets are present in a figure but only a few are actually in the visible extent. (e.g. during pan-zoom etc.)

This PR addresses this issue by evaluating the data-bbox on DSArtist initialization and then using it to fast-exit DSArtist.make_image() in case no data is in the extent.

[philippjfr]

Looks great to me!

[codecov]

Codecov Report

Attention: Patch coverage is 85.71429% with 1 line in your changes missing coverage. Please review.

Project coverage is 90.30%. Comparing base (7f31d04) to head (33e333e).

Files	Patch %	Lines
datashader/mpl_ext.py	85.71%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1345      +/-   ##
==========================================
- Coverage   90.31%   90.30%   -0.01%     
==========================================
  Files          92       92              
  Lines       18579    18582       +3     
==========================================
+ Hits        16779    16781       +2     
- Misses       1800     1801       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[raphaelquast]

Awesome! Thanks for the really quick response!

[jbednar]

Thanks so much for the contribution! I can't spot any reason not to merge this.

[raphaelquast]

Nice! Thanks for having a look!

As i see it, the only drawback that might be relevant is that coordinate limits are now always evaluated while before it was possible to circuumvent the evaluation by passing explicit values for x_range and y_range.

[jbednar]

Seems like an OK tradeoff to me.