feat(evals): publish multimodal_doc parser_compare benchmark + n=171 report

Adds the full parser_compare experiment for the multimodal_doc suite:
six arms compared on 30 PDFs / 171 questions from MMLongBench-Doc with
anthropic/claude-sonnet-4.5 across the board.

Source code:
- core/parsers/{azure_di,llamacloud,pdf_pages}.py: direct parser SDK
  callers (Azure Document Intelligence prebuilt-read/layout, LlamaParse
  parse_page_with_llm/parse_page_with_agent) used by the LC arms,
  bypassing the SurfSense backend so each (basic/premium) extraction
  is a clean A/B independent of backend ETL routing.
- suites/multimodal_doc/parser_compare/{ingest,runner,prompt}.py:
  six-arm benchmark (native_pdf, azure_basic_lc, azure_premium_lc,
  llamacloud_basic_lc, llamacloud_premium_lc, surfsense_agentic) with
  byte-identical prompts per question, deterministic grader, Wilson
  CIs, and the per-page preprocessing tariff cost overlay.

Reproducibility:
- pyproject.toml + uv.lock pin pypdf, azure-ai-documentintelligence,
  llama-cloud-services as new deps.
- .env.example documents the AZURE_DI_* and LLAMA_CLOUD_API_KEY env
  vars now required for parser_compare.
- 12 analysis scripts under scripts/: retry pass with exponential
  backoff, post-retry accuracy merge, McNemar / latency / per-PDF
  stats, context-overflow hypothesis test, etc. Each produces one
  number cited by the blog report.

Citation surface:
- reports/blog/multimodal_doc_parser_compare_n171_report.md: 1219-line
  technical writeup (16 sections) covering headline accuracy, per-format
  accuracy, McNemar pairwise significance, latency / token / per-PDF
  distributions, error analysis, retry experiment, post-retry final
  accuracy, cost amortization model with closed-form derivation, threats
  to validity, and reproducibility appendix.
- data/multimodal_doc/runs/2026-05-14T00-53-19Z/parser_compare/{raw,
  raw_retries,raw_post_retry}.jsonl + run_artifact.json + retry summary
  whitelisted via data/.gitignore as the verifiable numbers source.

Gitignore:
- ignore logs_*.txt + retry_run.log; structured artifacts cover the
  citation surface, debug logs are noise.
- data/.gitignore default-ignores everything, whitelists the n=171 run
  artifacts only (parser manifest left ignored to avoid leaking local
  Windows usernames in absolute paths; manifest is fully regenerable
  via 'ingest multimodal_doc parser_compare').
- reports/.gitignore now whitelists hand-curated reports/blog/.

Also retires the abandoned CRAG Task 3 implementation (download script,
streaming Task 3 ingest, CragTask3Benchmark + tests) and trims the
runner / ingest module APIs to match.

Co-authored-by: Cursor <cursoragent@cursor.com>

surfsense_evals/scripts/inspect_first30.py

@@ -0,0 +1,59 @@
+"""Inspect what the first 30 MMLongBench-Doc PDFs would look like for scoping.
+
+Run from surfsense_evals/ root via:
+    python scripts/inspect_first30.py
+
+Prints which docs are already ingested (existing 5), which are new (25 to
+upload), how many questions cover those 30 PDFs, and the answerable /
+unanswerable + format mix.
+"""
+
+from __future__ import annotations
+
+import json
+from collections import Counter
+from pathlib import Path
+
+
+def main() -> None:
+    qpath = Path("data/multimodal_doc/mmlongbench/questions.jsonl")
+    lines = qpath.read_text(encoding="utf-8").splitlines()
+    rows = [json.loads(line) for line in lines if line.strip()]
+
+    docs_by_id = sorted({r["doc_id"] for r in rows})
+    first30 = docs_by_id[:30]
+    existing5 = {
+        "05-03-18-political-release.pdf",
+        "0b85477387a9d0cc33fca0f4becaa0e5.pdf",
+        "0e94b4197b10096b1f4c699701570fbf.pdf",
+        "11-21-16-Updated-Post-Election-Release.pdf",
+        "12-15-15-ISIS-and-terrorism-release-final.pdf",
+    }
+    new25 = [d for d in first30 if d not in existing5]
+    print(
+        f"first 30 docs (alphabetical) — {len(new25)} new, "
+        f"{len(first30) - len(new25)} already in SurfSense"
+    )
+
+    qs_in_30 = [r for r in rows if r["doc_id"] in set(first30)]
+    fmts = Counter((r.get("answer_format") or "").lower() for r in qs_in_30)
+    answerable = sum(v for k, v in fmts.items() if k != "none")
+    unanswerable = fmts.get("none", 0)
+
+    print(
+        f"questions covering first 30 docs: total={len(qs_in_30)}  "
+        f"answerable={answerable}  unanswerable={unanswerable}"
+    )
+    print(
+        f"avg Qs/PDF: {len(qs_in_30) / 30:.1f}  "
+        f"answerable/PDF: {answerable / 30:.1f}"
+    )
+    print(f"format mix in scope: {dict(fmts)}")
+    print()
+    print("25 new PDFs to ingest:")
+    for d in new25:
+        print(f"  - {d}")
+
+
+if __name__ == "__main__":
+    main()