boolean[] backed by i8: 30% sieve speedup

[cs01]

Summary

boolean[] now uses 1-byte i8 elements (via %Uint8Array struct) instead of 8-byte double elements. For a 1M-element boolean array, this reduces the working set from 8MB to 1MB — fitting in L2 cache instead of spilling to L3.

Measured: 30% speedup on prime sieve benchmark (174ms → 247ms baseline, 100 iterations of 1M-element sieve)

Changes

• Compiler internals first: converted 6 compiler-internal boolean[] uses (outputIsTerminator, isConst, isDemoted) to number[] so the optimization doesn't affect self-hosting
• Type system: boolean[] maps to %Uint8Array* instead of %Array*
• Type inference: isUint8ArrayExpression detects boolean[] variables; isArrayExpression no longer matches boolean[]
• Variable allocator: boolean[] allocates as SymbolKind_Uint8Array
• Array literals: boolean array literals create %Uint8Array structs with i8 data
• Array push: boolean push stores i8 (0/1) instead of double (0.0/1.0)
• Index read/write: reuses existing Uint8Array paths (i8↔double conversion already works)

Test plan

• npm run verify:quick passes (tests + self-hosting Stage 1)
• All 4 boolean array test fixtures pass (basic, push, index-assign, prime sieve)
• Self-hosting unaffected (compiler internals use number[] now)
• 30% sieve speedup measured locally

[github-actions]

Benchmark Results (Linux x86-64)

Benchmark	C	ChadScript	Go	Node	Bun	Place
Binary Trees	1.475s	1.180s	2.676s	1.161s	0.968s	🥉
Cold Start	0.8ms	0.8ms	1.2ms	23.0ms	8.8ms	🥇
Fibonacci	0.814s	1.473s	1.560s	3.162s	1.946s	🥈
File I/O	0.116s	0.091s	0.087s	0.200s	0.173s	🥈
JSON Parse/Stringify	0.004s	0.005s	0.017s	0.015s	0.007s	🥈
Matrix Multiply	0.487s	0.695s	0.618s	0.362s	0.351s	#5
Monte Carlo Pi	0.388s	0.408s	0.404s	2.242s	6.030s	🥉
N-Body Simulation	1.665s	2.120s	2.205s	2.395s	3.255s	🥈
Quicksort	0.213s	0.245s	0.213s	0.261s	0.226s	#4
SQLite	0.342s	0.400s	—	0.419s	0.401s	🥈
Sieve of Eratosthenes	0.015s	0.028s	0.020s	0.039s	0.036s	🥉
String Manipulation	0.008s	0.044s	0.016s	0.035s	0.027s	#5

CLI Tool Benchmarks

Benchmark	ChadScript	grep	node	xxd	Place
Hex Dump	0.422s	—	0.918s	0.128s	🥈
Recursive Grep	0.018s	0.009s	0.093s	—	🥈