Missed optimization: fold sdiv exact by constant + icmp ugt into scaled unsigned bound check

[Leo0506]

define i1 @src(i64 %arg0) {
  %v0 = sdiv exact i64 %arg0, 12
  %v1 = icmp ugt i64 %v0, 12
  ret i1 %v1
}

define i1 @tgt(i64 %arg0) {
  %v0 = icmp ugt i64 %arg0, 144
  ret i1 %v0
}

alive2: https://alive2.llvm.org/ce/z/AtXTSY
godbolt: https://godbolt.org/z/3x9eePPjf
Pattern found in: https://github.com/dtcxzyw/llvm-opt-benchmark/blob/main/bench/llvm/optimized/ASTUnit.ll

[nikic]

Related: #76439