Subr-flattened charstrings are temporarily re-encoded in ByteStrBuff during "plan" phase, then copied to hb_serialize_context_t during "write" phase

CSOpSet may callback opcode processing "virtual" functions via CRTP

Numer struct may store a value as fixed optionally in addition to int and float

along with CFF2 charstring interpreter
factored out common code between CFF1 & CFF2 to CSInterpreter
moved fetch_op from Interpreter to InterpEnv
misc code clean up & bug fixes

to avoid confusion with OT::Index

along with HB_OT_TAG_cff1

"CFF" is used for shared structs & sources between CFF1 & CFF2
except OT:cff is for CFF version 1 table

Unreferenced subroutines are nullified (not removed) in a subset font

Separate Interpreter struct from hb-ot-cff-common-private.hh in a set of four header files, augmented for CharString (CS):
hb-cff-interp-common-private.hh
hb-cff-interp-dict-common-private.hh
hb-cff-interp-cs-common-private.hh
hb-cff-interp-cs.hh

Interpreter runtime is separated off as a new struct InterpEnv sub-classed differently for Dict and CharString (CS)

Add HB_USE_ATEXIT guard around free_static_C_locale missed by previous
commit.

Add HB_USE_ATEXIT macros around free_static function definitions to
avoid unused function compiler warnings/errors.

Users don't expect a new reference returned from a get() function.
Indeed, all users of that API I foud where NOT destroying the reference.
Just change the implementations to NOT return a reference.

This applies to the following APIs:
hb_unicode_funcs_get_default()
hb_glib_get_unicode_funcs()
hb_icu_get_unicode_funcs()

Fixes https://github.com/harfbuzz/harfbuzz/issues/1134

Fixes https://github.com/harfbuzz/harfbuzz/issues/1129

We've had the proper barriers already.  No need for more with every access.