Processing files sometimes gets stuck

[Chetic]

Processing files 22.3% (747/3357) was stuck on a PDF files for several days. It said "Loading file" with a frozen braille spinner when I got back after the weekend.
No clues on why it's stuck as far as I can tell.

[Chetic]

Also saw the braille spinner freeze for a long time (>15 mins) during "Splitting into sections". But this could just be that it takes a long time. Not sure.

[Chetic]

Fixed in 141ca1e. The root cause was that the heartbeat thread's timeout exception only broke its own loop — it never interrupted the main processing thread.

What changed:

1. Hard timeout wrappers for all file types — _load_data_with_timeout() (PDF/MD/TXT) and _split_docx_with_timeout() (DOCX/DOC) run file processing in a worker thread with future.result(timeout=...), so a hang in PyPDF or python-docx is caught regardless of internal timeout checks.

2. Main-thread timeout detection — Added check_timeout() and remaining_seconds() to FileProcessingContext. The main thread now checks both the _timeout_event (set by heartbeat) and wall-clock time, so it can raise FileProcessingTimeoutError itself rather than relying on the heartbeat thread.

3. PDF heading extraction fix — _extract_pdf_headings_from_outline() now caches page text lengths (reducing O(headings × pages) extract_text() calls to O(pages)), checks timeout_event and a wall-clock safety net per heading iteration, and returns partial results on timeout.

4. Unified loading paths — Sequential mode now uses load_files_parallel(max_workers=1) so future.result(timeout=per_file_timeout + 10) always acts as an outer hard timeout, closing the gap where sequential mode had no outer timeout at all.