refactor(codegen): migrate map-allocator to typeOf()

[cs01]

Summary

• Migrates 2 this.ctx.resolveExpressionType(stmt.value) call sites in map-allocator.ts (Map type inference at line 108, Set type inference at line 279) to the canonical this.ctx.typeOf(stmt.value).
• Same pattern as PR refactor(codegen): migrate array-allocator to typeof() (phase-e step 5, pr2) #575 (array-allocator), refactor(codegen): migrate class-allocator to typeof() (phase-e step 5, pr1) #576 (class-allocator), refactor(codegen): migrate 4 llvm-generator resolveExpressionType sites to typeOf() #583 (llvm-generator). Hits the annotator cache first, falls through to the resolver for context-dependent types.
• Widens MapAllocatorContext interface to expose typeOf instead of resolveExpressionType.

User-facing effect

• No behavior change. Continues Phase-E "typed AST" convergence — every codegen site reads types through one canonical lookup. Eliminates "two codegen sites asked the resolver at different times and got different answers" silent-wrong bugs for Map/Set allocation.

Test plan

• npm run verify — full tests + 3-stage self-host green locally
• CI green

[github-actions]

Benchmark Results (Linux x86-64)

Benchmark	C	ChadScript	Go	Node	Place
Cold Start	0.8ms	0.8ms	1.3ms	28.0ms	🥇
Fibonacci	0.814s	0.763s	1.561s	3.162s	🥇
Hash Map Lookup	0.101s	0.063s	0.092s	0.131s	🥇
Binary Trees	1.692s	1.332s	2.735s	1.223s	🥈
File I/O	0.118s	0.093s	0.087s	0.206s	🥈
JSON Parse/Stringify	0.035s	0.053s	0.177s	0.138s	🥈
N-Body Simulation	1.670s	2.123s	2.202s	2.403s	🥈
Regex Match	0.016s	0.005s	0.021s	0.004s	🥈
SQLite	0.052s	0.373s	0.497s	0.435s	🥈
Monte Carlo Pi	0.389s	0.410s	0.405s	2.249s	🥉
Quicksort	0.214s	0.246s	0.213s	0.263s	🥉
Sieve of Eratosthenes	0.017s	0.027s	0.018s	0.038s	🥉
String Manipulation	0.008s	0.019s	0.016s	0.037s	🥉
Matrix Multiply	0.436s	0.676s	0.614s	0.369s	#4

CLI Tool Benchmarks

Benchmark	ChadScript	grep	node	xxd	Place
Hex Dump	0.567s	—	0.979s	0.129s	🥈
Recursive Grep	0.020s	0.010s	0.103s	—	🥈