t27b: @intCast as an array index or a slice bound takes Zig's usize result type (specs/numeric/ocp_mx.t27) #7902

Part of #6063 (t27b coverage), item C10 of #6488.

What ocp_mx.t27 needs

specs/numeric/ocp_mx.t27 passes the reference (49 of 49 tests) and t27b refuses it: t27b test --blockers (master bbb77f0) names ExprCall(@intCast) at 42 sites, every one an array index, as in POW2[@intCast(e - 1)] and KIND_TOP[@intCast(kind)]. #7627 lowers @intCast only where the context names an integer type (@as, a typed binding, a parameter, a return). An index is not one of those contexts in t27b today, so the call has no result type there and is refused.

What the reference does

gen-zig prints POW2[@intCast(e - 1)] as written. In Zig 0.16 an index, and each bound of a slice, has the result type usize, so a[@intCast(k)] converts k exactly as @as(usize, @intCast(k)) would:

• an i64 or an i32 operand is checked, and a negative value is a safety panic in a Debug build (a failed test);
• a u8 or a u32 widens; a usize is kept;
• a literal folds, and a negative literal is a compile error.

The index is then bounds-checked as any index is.

Plan

• specs/tri/t27b/int_cast_plan.t27 names the result type an index and a slice bound give (INDEX_TYPE, usize). Its tests show plan's answers there: an i64 is checked and traps when negative, a u8 widens, a usize is kept, a literal folds or is refused.
• The glue lowers an @intCast index or slice bound through the existing int_cast with that type. It adds no new decision.
• A conformance spec in ocp_mx's shapes: specs/tri/t27b/conformance/int_cast_index.t27. The trap case (a negative index) joins cli/t27b/tests/int_cast_traps.t27.

Acceptance criteria

• t27b test --blockers specs/numeric/ocp_mx.t27 names no ExprCall(@intCast).
• The conformance spec passes under t27c test-report with 0 vacuous tests, and the same tests pass under t27b.
• The lab corpus receipt compare gives IMPROVED_ONLY, EQUIVALENT or NEUTRAL.

User Scenarios

• Given specs/numeric/ocp_mx.t27 contains array index expressions using @intCast such as POW2[@intCast(e - 1)], when t27b test --blockers specs/numeric/ocp_mx.t27 runs, then it names no ExprCall(@intCast) blockers.
• Given a conformance spec specs/tri/t27b/conformance/int_cast_index.t27 covering index and slice-bound @intCast with i64, u8, usize, and literal operands, when t27c test-report runs that spec, then it passes with 0 vacuous tests and the same tests pass under t27b.
• Given a negative i64 index used as @intCast in an array access, when the generated Zig runs in a Debug build, then it triggers a safety panic (trap) consistent with cli/t27b/tests/int_cast_traps.t27.

Requirements

• FR-001: The t27b lowering for @intCast when used as an array index or slice bound MUST assign the result type usize (the INDEX_TYPE from specs/tri/t27b/int_cast_plan.t27).
• FR-002: The lowering MUST reuse the existing int_cast lowering path with that usize result type; no new lowering decision logic is introduced.
• FR-003: The conformance spec specs/tri/t27b/conformance/int_cast_index.t27 MUST cover widening (u8→usize), preservation (usize→usize), checked conversion with trap on negative (i64→usize), literal folding, and negative literal rejection.
• FR-004: The trap behavior for negative index values MUST be exercised by cli/t27b/tests/int_cast_traps.t27 and pass under t27b.
• FR-005: After the change, t27b test --blockers specs/numeric/ocp_mx.t27 MUST report zero ExprCall(@intCast) sites.

Boundary

• specs/numeric/ocp_mx.t27
• specs/tri/t27b/int_cast_plan.t27
• specs/tri/t27b/conformance/int_cast_index.t27
• cli/t27b/tests/int_cast_traps.t27