[llvm] [DecoderEmitter] Support for DecodeOrder and `resolve-conflicts-try-all` (PR #157948)

[Rahul Joshi via llvm-commits]

jurahul wrote:

Thanks, yeah, the code in question seems to help AArch64 in eliminating the fallback namesapce, so in that sense its useful upstream. And as it happens, I think it will/can be made to work for my downstream case. And since it kicks in at the point we previously declared conflict, it should not regress any existing code and can then be gradually adopted. 

Note that currently the encodings in AllUnfilteredUnknown are not forced to be incomplete, just that they can be.

For the 0x/x0 case, they will be, we will still call it a conflict since there is no single instruction with all unfiltered bits unknown, so the new code will not kick in.

https://github.com/llvm/llvm-project/pull/157948