Fix query indexing at the cache boundary

[fallintoplace]

What

• Fix query indexing at the cache boundary.

Why

• With one cached result, get_item(1) could raise IndexError while more results exist.

Implementation

• Use a strict bounds check in both accessors.
• Cover cached, uncached, and out-of-range indexes in async and sync queries.

---

Note

Low Risk
Small bounds-check change in query indexing with targeted tests; behavior fix for partial caches, no auth or data-model changes.

Overview
Fixes off-by-one cache hits when indexing a FindQuery that only has a partial result set loaded (e.g. after execute(exhaust_results=False)).

Both sync __getitem__ and async get_item now treat the cache as usable only when len(_model_cache) > item instead of >= item. With a single cached row, index 1 no longer reads past the cache and raises IndexError; it clones the query with offset=1, limit=1 and fetches the next document.

Adds tests/test_find_query_cache.py to cover cached vs uncached indexes and out-of-range access for async get_item and sync [].

^{Reviewed by Cursor Bugbot for commit 2e2d779. Bugbot is set up for automated code reviews on this repo. Configure here.}