[llvm] [AMDGPU] Do not use 64-bit literals with VOP3* encodings (PR #156602)

[Jay Foad via llvm-commits]

jayfoad wrote:

> Do uses of `GCNSubtarget::has64BitLiterals` also require extra gating? Or is that one already fine, as is?

It depends on what they're doing, but there is nothing fundamentally wrong with calling `GCNSubtarget::has64BitLiterals`.

https://github.com/llvm/llvm-project/pull/156602