Limit chunked ingest chord fan-out to respect parser concurrency limits

[JSv4]

Motivation

• The non-eager chunked ingest path previously built a Celery chord over every prepared chunk and thus bypassed the parser's max_concurrent_chunks throttling, allowing an authenticated uploader to enqueue an unbounded number of chunk tasks and impact availability.

Description

• Enforce the parser concurrency limit in the non-eager branch by only creating the Celery chord when len(chunk_inputs) <= parser_instance.max_concurrent_chunks in opencontractserver/tasks/doc_tasks.py.
• When the chunk count exceeds max_concurrent_chunks, fall back to the existing in-process process_document path so the synchronous dispatcher and its thread-pool throttling still control parser-service concurrency.
• Add a regression test test_ingest_doc_large_pdf_falls_back_when_chunk_count_exceeds_limit in opencontractserver/tests/test_chunk_tasks.py that asserts oversized chunk sets do not call self.replace(...) / dispatch a chord and instead use the synchronous path.

Testing

• Static sanity check: compiled changed files with python -m py_compile opencontractserver/tasks/doc_tasks.py opencontractserver/tests/test_chunk_tasks.py (succeeded).
• Django test run attempted: python manage.py test opencontractserver.tests.test_chunk_tasks --keepdb (blocked in this environment due to missing django dependency, so tests could not be executed end-to-end here).

---

Codex Task

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!