Fallback hhea's ascender/descender to OS2

Reinstates previous logic, even if it might be unnecessary.

I believe that was a try to use one approach as fallback to other.  But
felt wrong.  Just believe what's in OS/2 table to be correct.

Fixes https://github.com/harfbuzz/harfbuzz/pull/1432

This reverts commit 1c8654ea.

Let's see if 576065b4 has fixed it

This reverts commit b8e90ca1.

Works now.

This is its failure https://circleci.com/gh/harfbuzz/harfbuzz/99864

Trying to fix like ee05627a, interestingly, makes the bot and the
others to fail like this https://circleci.com/gh/harfbuzz/harfbuzz/99841

Interestingly all of the bots disagreed with the change and the complain is... weird.

Decided to apply is we did the same on other places however this won't
fix all of its complains

Fixes #1834 at least till fix of #1829

Happens when compiled with -std=c++2a, the fix just makes the captures explicit to resolve the issue. Just adding this in addition to = doesn't work in C++11.

src/hb-ot-layout-gpos-table.hh:737:18: warning: implicit capture of 'this' with a capture default of '=' is deprecated [-Wdeprecated-this-capture]
              { return (this+_).intersects (glyphs, valueFormat); })
                        ^
src/hb-ot-layout-gpos-table.hh:736:16: note: add an explicit capture of 'this' to capture '*this' by reference
    | hb_map ([=] (const OffsetTo<PairSet> &_)
               ^
                , this

This reverts commit e4e74c27.

See https://github.com/harfbuzz/harfbuzz/issues/1829

Now is flaky let's just increase and maybe investigate later

And enable more tests able to trig the issue.

This reverts commit 62e60322 since it
breaks building HarfBuzz as part of Chromium.

Fixes https://github.com/harfbuzz/harfbuzz/issues/1821.

Yoohoo.  This seems to be precise enough!  Let's see if it sticks.
I'm asking Dominik to run this in Chrome test suite and report.

Fixes https://github.com/harfbuzz/harfbuzz/issues/1801

Part of https://github.com/harfbuzz/harfbuzz/issues/1801

Part of https://github.com/harfbuzz/harfbuzz/issues/1801

The assumption that compiler optimizes "upem/2" to a shift only
works if upem is unsigned...  Anyway, spoon-feed the compiler.

Fixes https://github.com/harfbuzz/harfbuzz/issues/1818

Just for those that are normally built into libharfbuzz itself.

Part of https://github.com/harfbuzz/harfbuzz/issues/1809

Part of https://github.com/harfbuzz/harfbuzz/issues/1809