fix(json): preserve Unicode boundaries in lexer errors

[bobzhang]

JSON lexer errors on supplementary Unicode characters can point at the trailing UTF-16 surrogate. For example, [true😀] reports column 6 instead of column 5, and numeric scanning can rewind into the middle of a surrogate pair.

Rewind by the consumed character's UTF-16 length and correct the U+FFFF BMP boundary in expect_char. Add regression coverage for separators, property names, numeric states, and the BMP boundary. Clarify that surrogate-escape validation does not validate raw UTF-16 input.

Validation:

• moon check --deny-warn
• moon test json --target all
• moon info; generated interfaces unchanged

[coveralls]

Coverage Report for CI Build 6626

Coverage increased (+0.004%) to 89.318%

Details

• Coverage increased (+0.004%) from the base build.
• Patch coverage: 19 of 19 lines across 3 files are fully covered (100%).
• No coverage regressions found.

Uncovered Changes

No uncovered changes found.

Coverage Regressions

No coverage regressions found.

---

Coverage Stats

Relevant Lines:	18414
Covered Lines:	16447
Line Coverage:	89.32%
Coverage Strength:	272979.92 hits per line

---

💛 - Coveralls