feat(stdlib): graduate std-llm/std-json to 1.0.0; enhance stdlib with new built-ins

[escapeboy]

Summary

• Graduates std-llm and std-json from 0.1.0 to 1.0.0 (4-week API stability window waived per operator decision); both packages are now LTS-protected from v1.3.0.
• Enhances std-json/src/core.ax with a real json_array(List<String>) -> String using __builtin_string_join (replacing the manual-join stub).
• Enhances std-validation/src/core.ax: wires string_length to __builtin_string_len (was hardcoded 0) and adds validate_contains, validate_starts_with, validate_ends_with using the new string built-ins.
• Updates CHANGELOG.md, docs/stdlib-graduation-tracker.md, and docs/lts.md to reflect the graduations.

Test plan

• cargo test -p boruna-tooling — 138 passed, 0 failed (includes test_std_llm_runs, test_std_json_runs, test_stdlib_all_compile_and_run, test_stdlib_determinism)

🤖 Generated with Claude Code

[github-actions]

Bench compare

Threshold for regression: ≥ 10% slower mean.

Benchmark	Mean change	99% CI
compile_crud_admin_template	-1.14%	[-1.34%, -0.90%]
compile_medium_program	-2.65%	[-2.93%, -2.26%]
compile_small_program	-10.51%	[-10.91%, -10.04%]
evidence_build_5_steps	+1.69%	[-0.46%, +3.92%]
evidence_build_empty	-0.75%	[-1.16%, -0.28%]
evidence_verify_10_steps	-1.12%	[-2.19%, +0.17%]
evidence_verify_5_steps	+0.64%	[-3.25%, +3.67%]
vm_call_dispatch_loop/iters=1000	-3.48%	[-3.83%, -2.95%]
vm_call_dispatch_loop/iters=10000	-2.94%	[-3.04%, -2.83%]
vm_pure_loop/iters=1000	-3.81%	[-3.99%, -3.59%]
vm_pure_loop/iters=10000	+1.59%	[+1.46%, +1.71%]
vm_pure_loop/iters=100000	+2.70%	[+2.61%, +2.79%]
vm_record_loop/iters=1000	-0.67%	[-0.83%, -0.50%]
vm_record_loop/iters=10000	-0.91%	[-0.99%, -0.83%]

✅ No benchmark regressed past threshold.