refactor(ascii): drop the duplicate scalar decode loop in JS decode

[bobzhang]

What

encoding/ascii/decode.mbt carried the same scalar ASCII decode loop twice:
once as the private decode_scalar (gated #cfg(not(target="js"))) and once
open-coded inside the JS pub fn decode (#cfg(target="js")). The two bodies
were byte-identical — verified by direct comparison of the pre-edit file: the
12 loop lines inside the JS decode equal decode_scalar's body, including the
(_, _ as bytes) => raise Malformed(bytes) arm and the shared finish_string
tail.

Before: on JS the ASCII acceptance rule and the Malformed payload
construction existed in a second, hand-maintained copy, separated from
decode_scalar by ~60 lines of v128 code.
After: decode_scalar loses the cfg and keeps its #inline (it names no
@v128 API, unlike decode_v128), and the JS decode body becomes
decode_scalar(bytes). Net −12 duplicated lines, −1 #cfg; the JS public doc
comment stays on the public function.

Behavior is unchanged:

• The JS success string is identical — the same code path runs.
• On JS the Malformed payload is still the remaining BytesView starting at
  the first non-ASCII byte, still a view (not a copy), and its Debug rendering
  (Malformed(<BytesView: [0xff]>)) is unchanged.
• The non-JS optimized decode, its 64-byte crossover to decode_v128,
  decode_v128's error slicing and suppress_unused_v128_import_on_js are
  untouched.
• Public API unchanged: pkg.generated.mbti has no diff, and no test snapshot
  changed.

This follows the doctrine already used by the sibling encodings:
encoding/hex/decode.mbt, encoding/base64/decode.mbt and
encoding/utf16/decode.mbt each keep a single decode_scalar and have their
non-vector decode delegate to it (encoding/utf16/decode.mbt:213-219 is the
same JS decode → decode_scalar delegation).

Validation

Run on this exact tree, base a0f0fb87cc0b506364f818c84725152bffbdbb97:

• moon test encoding/ascii --target all → 15/15 wasm, 15/15 wasm-gc, 12/12 js,
  15/15 native, exit 0 — identical counts to the pre-edit baseline.
• moon check --target all --deny-warn → exit 0.
• moon test (full repository) → 7609 passed, 0 failed, exit 0.
• moon bundle --all → exit 0.
• moon info → exit 0, no .mbti changes.
• moon fmt → exit 0, file unchanged.

Changed paths: encoding/ascii/decode.mbt only (1 insertion, 13 deletions).

CI at the single post-publication snapshot (gh pr checks 4235, taken
immediately after opening): version-check, misc-check, stable-check
(ubuntu/windows/macos) and stable-native-opt-test (ubuntu/windows/macos) are
pending; nightly-check is skipping; Devin Review and
GitGuardian Security Checks pass. Pending is recorded as pending — no CI wait
or polling was performed.

[coveralls]

Coverage Report for CI Build 6690

Coverage increased (+0.03%) to 89.386%

Details

• Coverage increased (+0.03%) from the base build.
• Patch coverage: 1 uncovered change across 1 file (0 of 1 lines covered, 0.0%).
• No coverage regressions found.

Uncovered Changes

File	Changed	Covered	%
encoding/ascii/decode.mbt	1	0	0.0%

Coverage Regressions

No coverage regressions found.

---

Coverage Stats

Relevant Lines:	18353
Covered Lines:	16405
Line Coverage:	89.39%
Coverage Strength:	275825.15 hits per line

---

💛 - Coveralls