build the page object index once to avoid quadratic scans

### Problem

In the `pypdfium2` PDF backend, `PyPdfiumPageBackend._object_rects` / `has_content_in` re-enumerate every object on the page on each call. `has_content_in` is invoked per layout cluster (2-3 times each), so on a page with many objects the cost is O(clusters × objects), and text extraction via `get_text_bounded` is O(rects × chars). A page with a large number of path/text objects makes conversion pathologically slow.

There is also a pre-existing hazard: the old `_object_rects` was a generator that held the process-global `pypdfium2_lock` across its `yield`, so overlap comparisons ran while holding the lock.

### Proposed fix

Build the page object index (bboxes bucketed by type, already mapped to the display frame) once per page and reuse it across all callers; clear it when the native page closes. This makes the object walk O(objects) once per page and lets callers iterate cached pure-Python lists with the lock released. Output is unchanged.

A PR implementing this follows.

Start with PyPdfiumPageBackend._object_rects and has_content_in, then trace the native page close path. Build the per-page object index once, reuse its display-frame bounding-box lists without holding the pypdfium2_lock, and clear it when the page closes. Done means output is unchanged and repeated scans no longer re-enumerate page objects.

Written by the indexing model from the issue text.