test[llm-llamacpp]: bounded FuzzTest for stop-string and log sanitizers

[jpgaribotti]

🎯 What problem does this PR solve?

• forLogMessage, toLowerAscii, and matchesAnyStopString take attacker-controlled prompt / template / antiprompt text. Empty stop entries, huge fragments, and control bytes are the same untrusted-input class as the NMT header parsers.
• They are header-only and already unit-tested, but had no property coverage under ASan + LSan.

📝 How does it solve it?

• Add llm-string-fuzz over those three helpers with no LINK_FABRIC (full ASan + LeakSanitizer).
• Seed empty stops, truncated log fragments, and case-folded template tags so bounded CI actually reaches the interesting paths.
• Wire fuzz* scripts, the vcpkg fuzz feature, and Linux npm run fuzz after the coverage harvest in cpp-tests-llm.yml.
• Patch bump @qvac/llm-llamacpp 0.53.2 → 0.53.3. No public addon API changes.

🧪 How was it tested?

• Follows the classification-ggml Phase 0 template (docs/architecture/ADDON-FUZZING.md).
• Bounded mode is what Linux C++ CI will run.
• on-pr-llm-llamacpp.yml is pull_request_target and reads workflow YAML from main, so the new fuzz steps run after this lands, not on this PR itself.

[github-actions]

Review Status

Current Status: ❌ PENDING
Approvals so far: none

Pending reviews: Needs 1 Management or Team Lead, and 1 more from Management, Team Lead, or Member.