fix: memory and semaphore leaks

[thismanyboyfriends2]

I've noticed from using OFscraper a fair bit that it doesn't seem to handle force closes or ctrl-c particularly well, especially on retries. Sometimes when running ofscraper a few times concurrently it seems to totally freeze up, even struggling to return the --help dialog. I've often had to resort to restarting the OS in order to free up resources.

This PR is an attempt to clear up some of those issues:

• Semaphore deadlock (main_download.py, alt_download.py) — Early returns didn't release semaphores; moved acquire/release to try/finally at method entry
• Manager reuse bug (manager.py) — manager is None was comparison, not assignment; fixed to manager = None
• ThreadPoolExecutor never shutdown (exit.py) — Added closeThreadExecutor() with wait=True to exit handlers
• Daemon thread infinite loop (scraper.py) — Changed daemon=True to daemon=False, added join(timeout=5) in exception handlers
• Event loop cleanup failures (run_async.py) — Replaced bare except: with specific exception handlers; explicit RuntimeError catch for already-closed loops
• Database pool not shutdown on success (wrapper.py) — Added finally block with shutdown(wait=True) on all paths
• Semaphore double-release risk (sessionmanager.py) — Consolidated manual releases into single finally block
• Asyncio task exceptions uncaught (download.py, metadata.py) — Added exception callbacks and return_exceptions=True to gather calls

I've been running it against a few profiles locally and this appears to work much better now, but I'm aware its a lot of changes - and there aren't many tests I can verify these fixes with. Any feedback is welcome.

Thanks!

[thismanyboyfriends2]

Okay I've realised this issue is a little trickier than I first realised, after testing it seems to have introduced some other issues. Let me look into it and fix it...