[al13n321]

Followup for #295

1. Turns out clang considers _BitInt (e.g. <builtin-type> ::= DB <number> _) substitutable.
   • AFAICT, this contradicts https://itanium-cxx-abi.github.io/cxx-abi/abi.html#mangling-compression . But it would be useful in practice to be able to demangle such symbols, so this PR adds this ability.
   • Clang is probably currently the only compiler emitting these DB <number> _ substrings in the wild (gcc doesn't allow _BitInt in C++). And this behavior is consistent between clang mangler and demangler. So I'm guessing the ABI doc will eventually change to match clang, not the other way around. So it seems to make sense to make this behavior default. If you disagree, I can just add an option instead.
   • I checked all other <builtin-type> productions with llvm-cxxfilt, _BitInt and vendor extended type are the only ones that are substitutable.
2. There was a simple bug where I didn't consume the _ after expression in DB <expression> _. Oops. Added a test.
3. Also changed "signed _BitInt(<n>)" to "_BitInt(<n>)" to match llvm-cxxfilt output.
4. Along the way also made "vendor extended type" builtin types substitutable, per https://itanium-cxx-abi.github.io/cxx-abi/abi.html#mangling-compression

(The code ended up a little weird: BuiltinType can be either in Type or in TypeHandle; some variants are always in Type, others are always in TypeHandle. Maybe there should instead be two different enums: substitutable builtin type (in Type) and non-substitutable builtin type (in TypeHandle). Lmk if you would prefer that, and how to name those enums. On the other hand, the current implementation allows easily adding an option to control whether to add _BitInt to substitution table.)

[khuey]

Thanks!