[llvm] [VectorCombine] Generalize foldBitOpOfBitcasts to support more cast operations (PR #148350)

[Rahul Yadav via llvm-commits]

rhyadav wrote:

@RKSimon fixed the ci format issue as there was a trailing space, also updated tests:

   - Keep one float-to-int bitcast test as a negative test case
  - Tests now demonstrate the optimization applies to integer bitcasts:
    
    pls approve


https://github.com/llvm/llvm-project/pull/148350