perf: fast integer path in number-to-string for ~38% stringops speedup

[cs01]

User-visible win

Programs that build strings by concatenating numbers (e.g. `"item" + i`, `"row-" + id`, logging loops, CSV/JSON formatting) now run noticeably faster. On the `stringops` benchmark:

• Before: ~17-18 ms
• After: ~10-11 ms
• ~38% faster — now within ~1.7x of hand-written C (~6 ms), down from ~3x.

Any app that formats numbers into strings in a hot loop benefits.

Root cause

`sample` profiling of the stringops benchmark showed ~45% of runtime inside `snprintf` — specifically `__vfprintf` / `__Balloc_D2A` / `__d2b_D2A` / `localeconv_l`. Every `"item" + i` concatenation was calling `snprintf("%.15g", double)`, which on macOS/glibc goes through a slow locale-aware floating-point path even when the value is a small integer.

The old `convertNumberToString` codegen also emitted:

1. `alloca` + `snprintf`
2. `strlen` on the result
3. `cs_arena_alloc`
4. `strcpy` (another full scan)

Four passes over the buffer per number.

Fix

• Added `cs_num_to_str(double)` bridge in `c_bridges/string-ops-bridge.c`. It takes the common fast path for integer-valued doubles (JS-compatible: no trailing `.0`, `-0` → `"0"`, safe range check against ±2^53), doing a direct itoa into an arena buffer with a single memcpy — no snprintf, no locale, no strlen, no strcpy.
• Non-integers, NaN, and Infinity still fall through to `snprintf("%.15g", ...)` inside the bridge, preserving exact semantics.
• `convertNumberToString` in the codegen is now a single `call @cs_num_to_str` instead of 7 emitted instructions.

Validation

• `npm test` — all 774 tests pass
• `npm run verify:quick` — PASSED (TypeScript + Stage 0 + Stage 1 self-hosting + tests)
• Sanity-checked `fibonacci`, `sieve`, `fileio`, `json`, `matmul` benchmarks — no regressions
• `stringops` phase breakdown (Date.now probes), integer-concat loop:
  • Before: ~10.2 ms
  • After: ~2.4 ms (~4x on the phase)

[github-actions]

Benchmark Results (Linux x86-64)

Benchmark	C	ChadScript	Go	Node	Place
Binary Trees	1.390s	1.248s	2.654s	1.206s	🥈
Cold Start	0.8ms	0.9ms	1.2ms	28.0ms	🥈
Fibonacci	0.911s	0.908s	1.731s	3.365s	🥇
File I/O	0.091s	0.097s	0.092s	0.176s	🥉
JSON Parse/Stringify	0.004s	0.005s	0.016s	0.015s	🥈
Matrix Multiply	0.493s	0.518s	0.787s	0.376s	🥉
Monte Carlo Pi	0.439s	0.441s	0.458s	2.601s	🥈
N-Body Simulation	1.760s	2.271s	2.282s	2.392s	🥈
Quicksort	0.245s	0.283s	0.242s	0.300s	🥉
SQLite	0.327s	0.363s	—	0.412s	🥈
Sieve of Eratosthenes	0.014s	0.027s	0.020s	0.042s	🥉
String Manipulation	0.008s	0.026s	0.016s	0.038s	🥉

CLI Tool Benchmarks

Benchmark	ChadScript	grep	node	xxd	Place
Hex Dump	0.433s	—	0.924s	0.140s	🥈
Recursive Grep	0.020s	0.011s	0.101s	—	🥈