[mlir][Vector] Prevent AVX2 lowering for non-f32 transpose ops

The AVX2 lowering for transpose operations is only applicable to f32 vector types.

Reviewed By: aartbik

Differential Revision: https://reviews.llvm.org/D120427