Tabular LLM generation acceleration phase 1

[Paul Lizer (paullizer)]

Summary

• Adds Phase 1 additive tabular generation contract fields while keeping the existing export contract version stable for old-run resume compatibility.
• Adds legacy-off rollout gates for the later planner, compact protocol, checkpointing, rolling pool, retry, heartbeat, and circuit-breaker phases, and filters backend-only rollout settings from sanitized user settings payloads.
• Adds safe batch observability for queue wait, model latency, validation time, checkpoint time, token counts when available, and character counts otherwise; removes generated response previews from mismatch telemetry.
• Extends the focused tabular scale suite with deterministic fake model/storage harnesses for out-of-order completions, usage counts, validation mismatches, and storage failures.
• Documents Phase 1 baseline contracts and validation status. Refs Scale per-row tabular analysis and generated CSV exports to 30,000+ rows #1031.

Validation

• Passed: python -m py_compile application\single_app\functions_tabular_generated_exports.py application\single_app\functions_settings.py application\single_app\config.py functional_tests\test_tabular_row_orchestration_scale.py
• Passed: python functional_tests\test_tabular_row_orchestration_scale.py
• Passed: python functional_tests\test_chat_multi_endpoint_notice_template_fallback.py
• Passed: git -c core.whitespace=blank-at-eol,blank-at-eof,space-before-tab,cr-at-eol diff --cached --check
• Blocked locally: python functional_tests\test_custom_logo_sanitization_fix.py did not reach sanitizer assertions because optional module olefile is not installed in this environment.

Notes

Live 300-row, 3,000-row, and 30,000-row model baselines still need to be captured in a configured deployment environment before later behavior rollout.