GB Stablehlo optimization

```
// CHECK-NEXT: %0 = stablehlo.iota dim = 0 : tensor<185xi64>
// CHECK-NEXT: %1 = stablehlo.add %0, %c_1 : tensor<185xi64>
// CHECK-NEXT: %2 = stablehlo.convert %1 : (tensor<185xi64>) -> tensor<185xf64>
// CHECK-NEXT: %3 = stablehlo.abs %2 : tensor<185xf64>
```

we shouldn't need to do the abs here as iota >= 0, as is iota + 1

The issue only shows a StableHLO CHECK-NEXT sequence; begin by locating the test or lowering pass that produces it and inspect the iota/add/convert/abs pattern. Done means the redundant abs is absent from the generated sequence and the relevant check passes.

Written by the indexing model from the issue text.