refactor(codegen): migrate array-allocator to typeof() (phase-e step 5, pr2)

[cs01]

Why this matters for users

Part of Phase E's type-resolver consolidation. Each migrated site is one less place where two codegen paths can disagree about an expression's type and silently produce wrong native code. No user-visible behavior change in this PR — it's a mechanical swap that routes a query through the annotator's pre-populated cache.

Change

ArrayAllocator.buildInterfaceElementInfo formerly called ctx.resolveExpressionTypeRich(expr) directly. Now calls ctx.typeOf(expr) — same semantics (returns ResolvedType | null), but hits the annotator cache first, only falling through to resolveExpressionTypeRich for expressions the annotator skipped (unknown-base).

Interface ArrayAllocatorContext updated accordingly.

Verification

Full npm run verify (tests + stage 2 self-host) green locally.

Scope

One call site. First consumer migration following the Step-5 annotator landing in #573. More migrations land in subsequent PRs.

[github-actions]

Benchmark Results (Linux x86-64)

Benchmark	C	ChadScript	Go	Node	Place
Binary Trees	1.644s	1.377s	2.762s	1.269s	🥈
Cold Start	1.0ms	0.8ms	1.3ms	30.2ms	🥇
Fibonacci	0.814s	0.766s	1.564s	3.173s	🥇
File I/O	0.119s	0.092s	0.090s	0.209s	🥈
JSON Parse/Stringify	0.004s	0.005s	0.017s	0.017s	🥈
Matrix Multiply	0.435s	1.005s	0.566s	0.358s	#4
Monte Carlo Pi	0.389s	0.410s	0.405s	2.249s	🥉
N-Body Simulation	1.667s	2.119s	2.210s	2.390s	🥈
Quicksort	0.215s	0.247s	0.214s	0.262s	🥉
SQLite	0.349s	0.368s	—	0.469s	🥈
Sieve of Eratosthenes	0.013s	0.028s	0.020s	0.038s	🥉
String Manipulation	0.008s	0.019s	0.016s	0.037s	🥉

CLI Tool Benchmarks

Benchmark	ChadScript	grep	node	xxd	Place
Hex Dump	0.563s	—	1.033s	0.129s	🥈
Recursive Grep	0.021s	0.011s	0.107s	—	🥈