fix: escape schema values in generated code

[mcollina]

Summary

• stop embedding schema references directly into generated comments, template literals, validator calls, and regular-expression literals
• encode complete generated error literals, including required-property errors
• escape JSON Pointer tokens and URI-fragment characters in generated validator references
• reject active schema identifiers containing block-comment terminators when Ajv source generation is enabled
• add regression coverage across normal, debug/restore, standalone, and inline-validator modes

Rationale

The README requires schema definitions to be trusted application code. This change preserves that trust model while ensuring schema-derived strings cannot accidentally change generated JavaScript or resolve to the wrong validator path.

Validation

• npm test — 514 tests passed; 100% statement, branch, function, and line coverage
• npm run lint
• targeted tests on Node.js 20, 22, 24, and 26
• npm pack --dry-run
• git diff --check

[Tony133]

This is inside the for (const key of Object.keys(...)) loop, so the RegExp gets constructed once per key per pattern on every serialization. Was that intentional, or worth hoisting it to module scope next to the generated functions?

[gurgunday]

lgtm