Part of https://github.com/harfbuzz/harfbuzz/issues/1553

Doesn't matter, but matches the description.

COLR table has one function: return layers for a glyph, and we expose exactly
that, so should just wire it through.  Also use sub_array() for verifiable
safety.

Also, BaseGlyphRecord's null object is enough.  We don't need to special-case
the not-found.

Implement has_data() for realz.

Not sure if MSVC would be unhappy about this.

Sanitize always comes just before data member definitions, so
it's easy to cross-check.

When a struct is plain old data with no references, etc, it's okay
to mark its members public.

...like UnsizedArrayOf<>.

This fixes a class of crasher bugs, mostly with color and AAT tables.  We
cannot use nullable offsets to varsized data that does not declare min_size,
because it's nost safe to use our fixed-size null pool for types that have
their size external.  So, use non_null'able offsets for these.

A further enhancement would be to make use of min_size in Null<> itself.
Will try that after.

Sorry for the noise, downstream custom builders.  Please adjust.

Bad bad bad bad code. Don't do that. If compiler's not happy, understand why.

* Implemented a bsearch on get_base_glyph_record
* Made get_color_record_argb actually work

That sanitization wasn't right, firstLayerIndex is an index from beginning of the Layer Records
not the table itself.