Array indexing in type checking for constexpr index does not match IR behavior

**Describe the bug**
The DSLX type system validation will error when a constexpr array index is out of bounds:

https://github.com/google/xls/blob/c59b6a21b8bc362f2899706bd74eb8ea7708644f/xls/dslx/type_system/deduce_utils.cc#L404-L410

However, as discussed in https://github.com/google/xls/discussions/1494, the IR documentation states:

> Any out-of-bounds indices idx_{i} are clamped to the maximum in bounds index for the respective dimension.

**To Reproduce**

```
fn foo(arr: u32[10]) -> u32 {
arr[10]
}

#[test]
fn foo_test() {
assert_eq(foo(u32[10]:[u32:1, ...]), u32:1);
}
```

yields the error:

> TypeInferenceError: uN[32][10] Index has a compile-time constant value 10 that is out of bounds of the array type.

Note that this behavior appears to be retained in TIv2:

https://github.com/google/xls/blob/c59b6a21b8bc362f2899706bd74eb8ea7708644f/xls/dslx/type_system_v2/typecheck_module_v2_test.cc#L735-L742