Thanks for the contribution! Before we can merge this, we need @Dnreikronos to sign the Fuel Labs Contributor License Agreement.

PR Summary

Medium Risk
Touches core expression lowering for all index reads; behavior change is narrow (extra loads for multi-level pointers) but mistakes could affect indexing correctness.

Overview
Fixes a compiler ICE when reading through an index on a ref mut array prefix (e.g. self[i] in impl ... for [T; N]). IR lowering left an extra level of indirection on the prefix compared to the reassignment path, which hit todo!() for typed-pointer array indexing.

compile_indexing now loads through nested pointer/reference layers until the prefix points at the real array or slice, then dispatches as before. Unsupported pointer shapes return an internal compile error instead of panicking.

The const-generics reassignment e2e test re-enables assert_eq(self[i], default) that was blocked on this bug.

^{Reviewed by Cursor Bugbot for commit b863ab9. Bugbot is set up for automated code reviews on this repo. Configure here.}

@ironcev, can we re-trigger the cla pipe here and on the other PRs that I have open?

Merging this PR will not alter performance

✅ 25 untouched benchmarks

_{Comparing Dnreikronos:fix/index-read-ref-mut-array-7602 (b863ab9) with master (ac933ff)}

bugbot run

@ironcev ready for a look now that CLA's green. The fix lives in compile_indexing: it loads through the nested pointer layers on a ref mut array prefix before dispatching, instead of hitting the todo!() for typed-pointer array indexing. I re-enabled the assert_eq(self[i], default) in the const-generics e2e test. Main thing to sanity-check is whether the extra loads are right across all the pointer shapes.