Add `InstructionBuilder::foo_imm_{u,s}` methods and signed/unsigned immediates to CLIF

[fitzgen]

Now that we're generating these helpers automatically, I wonder if it would make sense to have _imm_u and _imm_s variants? And maybe newtype the immediates (so we have SImm128 / UImm128 etc)? I guess my overall thought is that "historical semantics" carry some weight as far as back-compat goes but not too much, and as long as we're cleaning up we may as well clean up the interface too...

Originally posted by @cfallin in #13527 (comment)