* fix: skip page extraction when operation count exceeds 1M
Vector-heavy architectural PDFs can have 10-26M path operators per page,
causing ~60 GB allocation during per-op processing. The decode itself is
tolerable but the subsequent loop amplifies memory 2-5x with text items,
rects, paths, and state tracking.
Check operation count after Content::decode() and return empty extraction
for pages exceeding the limit, with a warning log.
* test: add unit test for excessive operations guard
Constructs a synthetic PDF with 1.1M path operators to verify
that pages exceeding the operation limit return empty extraction.
Pages with tens of thousands of vector-drawing rects (e.g. architectural
plans with 36k+ rects) caused cluster_rects and sub-rect deduplication to
spend 10-50s in O(n²) loops. No real table has thousands of cell rects,
so once a union-find component exceeds 2000 elements we skip further
comparisons, and skip the O(n²) sub-rect dedup entirely for such pages.
Benchmarks on 8 slow production PDFs show 10-40x speedups on vector-heavy
files (12s→0.3s) with no regression on other file types. All 342 tests
pass.
Roman Yurchak