[llvm] [AMDGPU] Only try DecoderTables for the current subtarget. NFCI. (PR #82992)

[Ivan Kosarev via llvm-commits]

kosarev wrote:

> If `DisassemblerEmitter` worked like that then I don't think we would need to use different `DecoderNamespace`s at all.

Yes, that's the point, wouldn't need multiple decoding attempts at all. I can imagine `DecoderNamespace` would still have its use cases where instructions to decode are completely unrelated.

https://github.com/llvm/llvm-project/pull/82992