[llvm] [AArch64] Fix type mismatch in bitconvert + vec_extract patterns (PR #183549)

[via llvm-commits]

================
@@ -8212,10 +8212,8 @@ defm : Neon_INS_elt_ext_pattern<v8i16, v4i16, v2i32, INSvi16lane, DUPi16, hsub,
 
 // bitcast of an extract
 // f32 bitcast(vector_extract(v4i32 src, 0)) -> EXTRACT_SUBREG(src)
-def : Pat<(f32 (bitconvert (i32 (vector_extract v16i8:$src, (i64 0))))),
----------------
Lukacma wrote:

So I removed it, because it is incorrect and I couldn't figure out how to test the fixed version. I can submit an untested fix, but since adding untested or insufficiently tested patterns caused this whole issue I am hesitant about adding it. If you think its fine though, I am happy to fix it instead

https://github.com/llvm/llvm-project/pull/183549