pyspy: idle and gil_only filtering happens too late

The pyspy collects all the traces for all the threads just to later drop them if not needed, this may consume significant cpu time.

Start by locating where pyspy collects thread traces and where the idle and gil_only filters are applied. Trace the current collection path to understand why unwanted traces are gathered first. Done means unnecessary traces are excluded earlier and the behavior is covered by the relevant existing tests, if available.

Written by the indexing model from the issue text.