* src/hb-cff-interp-dict-common.hh: Use ull for unsigned int64_t

The llu suffix does not work for older Visual Studio versions
(pre-2013), but ull works for all the compilers that we attempt to
support.

* test/api: Fix build on pre-C99 compilers

Ensure variables are declared at the top of the block.

* src/hb-dsalgs.hh: Add specialization for hb_is_signed<> for __int8

Pre-Visual Studio 2010 does not consider __int8 (which is typedef'ed to
int8_t) to be equivilant to signed char, so the compiler cannot find the
corresponding hb_is_signed<> specialization that is needed.

The interesting thing is unsigned __int8 is considered to be equivilant
to unsigned char, so as the other types (short, int, long) that we look
for here, so only the specialization for __int8 is added here.

This will fix builds on Visual Studio 2008 at least.

* Update hb-ot-cff-common.hh

* fix bug

* bummer fix wasn't hit. refix

* additional sanity check

* Added test cases for oss-fuzz issues 11805, 11806

* add fd index checks to subr subsetter

also added oss-fuzz test case

* undid SubrSubsetParam::is_valid

because already validated by SubrClosures.valid

* fix oss-fuzz 11690: substr.offset >= opStart

detect recursive subroutine call & handle as error

* fix build failure

* add minimized test case for oss-fuzz 11690

* removed asserts

Ensure variables are declared at the top of the block.

Also added an additional error check to avail ()

* guard against no subr access

* code tweak

* add minimized testcase for oss-fuzz 11670 (Null deference)

Happened after 11d2f49a
since now nClasses is unsigned int...

As run-tests.py already adds it

It is not visually noticeable but apparently affected by kern format2 correct implementation.
I should've checked CoreText result which can't as CircleCI outage.

Fixes https://github.com/harfbuzz/harfbuzz/pull/1346