[oxura]

Fixes #1612.

Counting a String no longer expands it into a character array or converts its non-scalar Chars at a text edge. The loop walks the existing internal Char widths; the head/tail decoder now accepts an offset, and both existing call sites pass zero. Extra counting storage changes from O(n) to O(1).

I compiled the same Bend program returning two million emoji, then measured three fresh Bun processes per build with /usr/bin/time -f %M. Every run printed 2000000. These are whole-process peak RSS medians, not an allocation-profiler estimate:

build	median peak RSS (KiB)
before	170988
after	56092

The mixed ASCII/astral/arbitrary-U32 regression, string_kit, char_any_code and string_unicode all pass on Bun and native C. Repo gate: 54/54.

Based on 7e120d5; compiler size is 63,618 → 63,696 (+78). I left this draft alongside #1606/#1610 while main remains at 64,092 against the permanent 64,000 cap. No cap changes, comment trimming or new dependencies.

Linux x86-64, Bun 1.4.3, clang 22.1.8; the mini-cluster/Metal/CUDA gates were not run here. AI-assisted patch.

[oxura]

This is JS-only performance work, so I’m withdrawing it following the reference-runtime policy in #1644.