[llvm] [TableGen][DecoderEmitter] Add option to emit type-specialized `decodeToMCInst` (PR #146593)

[Craig Topper via llvm-commits]

topperc wrote:

> The suggestion to use `std::bitset<>` seems reasonable. So we will use `uint_8/16/32/64_t` when possible, else fall back to `std::bitset<N>`. @topperc WDYT? Given that that will be an backward incompatible change, we still need a single bool option in `InstrInfo`, and once we have migrated all backends, we will plan to deprecate the templated code at some point in future? (giving enough time for downstream backends to migrate) Maybe that can discussed at a later point.

Can you do a field extract from a std::bitset?

https://github.com/llvm/llvm-project/pull/146593