The conversion was done with

for f in lib/**/baseline_images/**/*.svg; do
    gzip --best <"$f" >"$f"z; rm "$f"
done
sed -Ei 's/(extensions=.*)svg/\1svgz/g' **/*.py

(svgz is literally gzip'ed svg)

and manual fixes to the tests (in a separate commit).

The global size of all test svgs was 15Mb; it is 4Mb after compression.

Obviously that won't help with objects stuck in the git repository; instead, the main advantage is to make future addition of more svgs less problematic (the largest objects in the git repo are some svgs >5Mb (that have been deleted but are stuck in the history forever)), and also to prevent Github from showing svgs in the PR diff interface (although the later can also be achieved using gitattributes).

Alternative (or complementary to) #9215.

EDIT: Actually git appears to zlib-compress its objects (https://matthew-brett.github.io/curious-git/reading_git_objects.html) so this PR may not be that helpful wrt repo size...

PR Summary

PR Checklist

• Has Pytest style unit tests
• Code is PEP 8 compliant
• New features are documented, with examples if plot related
• Documentation is sphinx and numpydoc compliant
• Added an entry to doc/users/next_whats_new/ if major new feature (follow instructions in README.rst there)
• Documented in doc/api/api_changes.rst if API changed in a backward-incompatible way