[subset] zero glyf and loca memory before writing so pads are clean. Test just the part of maxp we care about

[subset] sketch out dropping in-glyf instructions. Sometimes yields differnet size glyphs than fonttools, possibly due to padding not being corrected

[subset] fix calculation of range shiftz. Should be 16 * len - searchRange not 16 * (len - searchRange).

Clean up of glyf subsetting.

I know it's bad form; I know I've said no to this multiple times...
But this is part of an experiment to see if we can make rolling HarfBuzz
into Chrome much faster. That will give us a lot more testing exposure
that currently Chrome does.

Doing this while those files are still considered ignored by our
gitignore rules; will see how that flies.

Some of our bots / compilers don't like it. Wasn't caught by bots because this
is only in debug code.

https://github.com/harfbuzz/harfbuzz/pull/809/commits/825ad378e0f0e6e6bb690894a0d5d25f198be21b

Fixes Chromium Android builds, compare
https://ci.chromium.org/buildbot/tryserver.chromium.android/linux_android_rel_ng/491787

Fixes #805

Fixes https://bugs.chromium.org/p/oss-fuzz/issues/detail?id=6022

Context needs to be called 'c'. The TRACE_* rely on that.

Previously adjusted logic could fail if nothing around was ligated,
and something moved.  This should take care of that.

CC https://github.com/harfbuzz/harfbuzz/issues/740

Add a continuous builder using fswatch