PR Summary

We've been having some internationalization issues caused by MPL not loading our international font, Microsoft YaHei. Part of the problem is that YaHei got turned in to a .TTC by a Windows update, which we rectified by overlayingthis PR. But we still weren't getting what we wanted because matplotlib was displaying using YaHei Light instead of regular YaHei, and it looked really bad.

We tracked the issue down to the fact that the font manager wasn't identifying the light and regular versions as having different weights, making them appear identical and leading to plots using whichever one was higher up in the list.

We rectified this issue by givingfont_manager.ttfFontProperty the ability to read Microsoft's SFNT tables. This allows matplotlib to correctly interpret the properties of a bunch of Windows default fonts including the one we were specifically interested in.

We just did this:

    sfnt2 = sfnt.get((1, 0, 0, 2), b'').decode('latin-1').lower() or \            sfnt.get((3, 1, 0x0409, 2), b'').decode('utf_16_be').lower()    sfnt4 = sfnt.get((1, 0, 0, 4), b'').decode('latin-1').lower() or \            sfnt.get((3, 1, 0x0409, 4), b'').decode('utf_16_be').lower()

and it started working (the utf_16_be lines are the new ones.)

I haven't written a test for this yet since, honestly, we're just wondering if there's some obvious reason that this is a bad idea and hasn't been done already. If the community think it's a worthwhile addition I can finish it up as a proper PR though.

PR Checklist

• Has Pytest style unit tests
• Code isFlake 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