Extend caching plugin to handle markdown and iptc metadata

[dsschult]

Turn markdown metadata in gallery.py into a property so it can be overridden.
Extend the caching plugin to handle markdown and iptc metadata (previously only handled exif).
Invalidate cache when a file timestamp changes, so we always get the latest data.
Cache file modification time lookups within a build to reduce load on filesystem.

Supersedes #440.

[codecov]

Codecov Report

Merging #443 (78ecbfe) into main (3b722ad) will increase coverage by 0.59%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main     #443      +/-   ##
==========================================
+ Coverage   87.33%   87.93%   +0.59%     
==========================================
  Files          23       23              
  Lines        1943     2014      +71     
==========================================
+ Hits         1697     1771      +74     
+ Misses        246      243       -3     

Impacted Files	Coverage Δ
sigal/gallery.py	91.28% <100.00%> (+0.32%)	⬆️
sigal/plugins/extended_caching.py	93.75% <100.00%> (+12.50%)	⬆️
sigal/utils.py	96.42% <100.00%> (+0.27%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 3b722ad...78ecbfe. Read the comment docs.

[corsmith]

I have an album with 150K pictures and 4K videos and use the extended_caching plugin. The 2.2 release takes approximately 18 minutes to rebuild the gallery with the plugin enabled (assuming no thumbnails to generate) and the cache file is approximately 23MB. Using the code in this PR it takes less than 5 minutes to rebuild the gallery with a cache size of 1.03GB (43x the previous size!). The pickle.load call in my setup alone takes 60 seconds of CPU time! While that is a little crazy the net effect is a 60% reduction in wall time to complete a full rebuild even with the previous extended_caching in 2.2. It is faster to parse the single file than to read 154K files for metadata across the filesystem as currently done in gallery.py -> class Media -> init -> self._get_metadata()

[saimn]

Sorry it seems I forgot to review this PR. I just had a look and it looks very good.
@dsschult - could you rebase it on main to make sure tests still run ?
If I had one comment about the code, maybe raw_metadata could be renamed to something like markdown_metadata or mkd_metadata ?

[dsschult]

@saimn should be good now

[saimn]

Excellent, thanks for the quick update !