[mlir] vector extract/insert identity canonicalization

[sri-cherukuri]

Trivial extract-insert identity pattern does is not folded by vector canonicalization passes

  %0 = arith.constant dense<0> : vector<2xi32>
  %1 = vector.extract %arg0[0] : i32 from vector<2xi32>
  %2 = vector.extract %arg0[1] : i32 from vector<2xi32>
  %3 = vector.insert %1, %0[0] : i32 into vector<2xi32>
  %4 = vector.insert %2, %3[1] : i32 into vector<2xi32>

%4 should be replaced with %arg0

https://godbolt.org/z/jY37h8T1Y

[llvmbot]

@llvm/issue-subscribers-mlir

Author: None (sri-cherukuri)

Trivial extract-insert copy does not folded by vector canonicalization patterns ``` %0 = arith.constant dense<0> : vector<2xi32> %1 = vector.extract %arg0[0] : i32 from vector<2xi32> %2 = vector.extract %arg0[1] : i32 from vector<2xi32> %3 = vector.insert %1, %0[0] : i32 into vector<2xi32> %4 = vector.insert %2, %3[1] : i32 into vector<2xi32> ``` `%4` should be replaced with `%arg0`

https://godbolt.org/z/jY37h8T1Y

[sri-cherukuri]

Seemingly resolved by #142944

[llvmbot]

@llvm/issue-subscribers-mlir-vectorops

Author: None (sri-cherukuri)

Trivial extract-insert identity pattern does is not folded by vector canonicalization passes ``` %0 = arith.constant dense<0> : vector<2xi32> %1 = vector.extract %arg0[0] : i32 from vector<2xi32> %2 = vector.extract %arg0[1] : i32 from vector<2xi32> %3 = vector.insert %1, %0[0] : i32 into vector<2xi32> %4 = vector.insert %2, %3[1] : i32 into vector<2xi32> ``` `%4` should be replaced with `%arg0`

https://godbolt.org/z/jY37h8T1Y