Avoids USHORT, SHORT, ULONG, LONG clashes with Windows API.

Gives me a good 10% speedup for the Devanagari test case.  Less so
for less lookup-intensive tests.

For the Devanagari test case, the false positive rate of the GSUB digest
is 4%.

If there is no GPOS, zero mark advances.

If there *is* GPOS and the shaper requests so, zero mark advances for
attached marks.

Fixes regression with Tibetan, where the font has GPOS, and marks a
glyph as mark where it shouldn't get zero advance.

This commit: a3313e54 broke MarkMarkPos
when one of the marks itself is a ligature.  That regressed 26 Tibetan
tests (up from zero!).  Fix that.  Tibetan back to zero.

And use it to speed up the hotspot by checking coverage directly in
the main loop, not 10 functions deep in.

Gives me a solid 20% boost with Indic test suite.  Less so for less
lookup-intensive scenarios.

Remove the "fast_path" hack from before.

This was broken as a result of 7b84c536.
As Khaled reported, MarkMark positioning was broken with glyphs
resulting from a MultipleSubst.  Fixed.  Test with the ALLAH character
in Amiri.

Should fix warnings generated when building with -Wunused-private-field.
Based on patch from Jonathan Kew.

Fixes more of Telugu, Kannada, and Oriya.

May break things (outside Indic...), but we cannot think of any font relying
on this immediately.

Untested.

This reverts commit 727135f3.

This is work-in-progress.  Didn't mean to push it out just yet.

Shaves another 3% for DejaVu Sans long Latin strings.

This is apparently what Uniscribe does.  Test case is:

  SEEN FATHA TEH ALEF

with Arabic Typesetting.  Originally reported by Khaled Hosny.

Introduced in 99c26957.
Broken mark-mark and mark-ligature stuff.

The spec doesn't say contextual matching should be done this way,
and AOTS doesn't do it either.  It was inherited from old HarfBuzz.
Remove it.

This is to incorporate it into GSUB/GPOS processing.

More to come, for APPLY, CLOSURE, etc.