Preserve cancellation during cached-content reads - #2481

Summary

Preserve task cancellation while reading cached content. _load_content() currently catches CancelledError with a bare except, returns None, and lets the cache request continue as though the content were missing. Catching Exception keeps ordinary filesystem/decode failures on the existing fallback path while allowing cancellation to propagate through the file and database cleanup.

List of files changed and why

• crawl4ai/async_database.py: catch ordinary exceptions when loading cached content.
• tests/unit/test_database_content_cancellation.py: cover real file reads/error fallbacks and cancellation during an established real FIFO read. The POSIX-only cancellation test skips on other platforms.

How Has This Been Tested?

On macOS / Python 3.14, the normally imported current package passes both committed regression tests (2 passed). A separate real AsyncDatabaseManager / SQLite cache-path check passes 12 checks, including acache_url / aget_cached_url, cancellation propagation, file and connection cleanup, and reuse of the connection-pool slot. Real file/FIFO controls also pass on Python 3.13 and 3.14.

Validation covers the cache-content and database path. The full project suite, browser/model execution, remote websites and Windows were not exercised. This does not make executor-backed filesystem operations interruptible in every phase.

Checklist

• Based on and targeting develop.
• Performed a self-review and added focused regression tests.
• Focused package tests and real SQLite/cache cleanup checks pass locally.
• No public API, database schema or documentation changes are required.