[OsamaAnsar]

Description of change

GbnfStringValue (used for enum / const string values and object property names) escaped only \, tab, CR, LF and " when building the grammar literal. Backspace, form feed and the other control characters U+0000 - U+001F were put into the literal raw. The model would then have to emit a raw control character, which is invalid inside a JSON string (and GbnfInsideStringChar excludes \x00-\x1F for ordinary strings), so JSON.parse of the output fails.

This escapes them the way JSON.stringify does: \b, \f, and \u00XX for the rest. Characters that are valid in JSON strings (DEL, non-ASCII, U+2028) are unchanged.

Before / after (real output of getGbnfGrammarForGbnfJsonSchema, control characters shown as <0xNN>):

{"enum": ["a<BS>b"]}      before: val0 ::= "\"a<0x08>b\""      after: val0 ::= "\"a\bb\""
{"enum": ["a<FF>b"]}      before: val0 ::= "\"a<0x0c>b\""      after: val0 ::= "\"a\fb\""
{"enum": ["a<0x01>b"]}    before: val0 ::= "\"a<0x01>b\""      after: val0 ::= "\"a\u0001b\""
{"const": "x<0x1f>"}      before: "\"x<0x1f>\""                after: "\"x\u001f\""
object property "k<BS>"   before: "\"k<0x08>\""                after: "\"k\b\""

Tests: new standalone test/standalone/utils/GbnfStringValue.test.ts (no native addon or model needed). It checks the escaped form for \b, \f, \u0000, \u0001, \u001f, that all 32 control characters match the JSON.stringify form and round-trip through JSON.parse, that the existing escapes are unchanged, and the full schema grammar for enum, const and property names. With only the source fix reverted 4 of the 6 tests fail; with the fix all 6 pass. tsc --noEmit and eslint on src/utils/gbnfJson and test/standalone/utils pass.

Not tested: the native addon could not be built in my environment, so the grammar was checked only as text (not compiled by llama.cpp, no sampling with a model), and LlamaGrammar.test.ts and the model-dependent tests were not run.

Pull-Request Checklist

• Code is up-to-date with the master branch
• npm run format to apply eslint formatting
• npm run test passes with this change - the new tests, tsc --noEmit and eslint pass; tests that need the native addon / models were not run
• This pull request links relevant issues as Fixes #0000 - N/A, no existing issue found
• There are new or updated unit tests validating the change
• Documentation has been updated to reflect this change - N/A, no API or usage change
• The new commits and pull request title follow conventions explained in pull request guidelines (PRs that do not follow this convention will not be merged)