improve the performance of STAC based back-end #972

I ran some tests to asses the wall-time and the total time spent in the database for different backends.

The STAC one is underperforming.

Likely candidates:

• Results seem come back 10 at a time. get_by_status and get_by_indices call client.search(...) without a limit, so the server's default page size applies We can consider increasing this one

• Counting downloads every job. count_by_status calls get_by_status, downloads every matching item over those pages, then counts them in Python.

• A full count runs every loop (job_db.count_by_status()). The f-string is built even when INFO logging is off, so each loop downloads the whole collection just to write a log line that never appears.

• Every save first lists all collections. persist() calls exists(), which pages through every collection on the server.

.