Update http cache rules to be more effective

[glensc]

Such overrides actually work now:

  policy:
    "api.trakt.tv/users/me": 12d
    # Trakt search urls, 304 works
    "api.trakt.tv/search/imdb/tt*?type=movie": "7d"
    "api.trakt.tv/search/imdb/tt*?type=show": "7d"
    "api.trakt.tv/search/tmdb/*?type=movie": "7d"
    "api.trakt.tv/search/tmdb/*?type=show": "7d"
    "api.trakt.tv/search/tvdb/*?type=show": "7d"
    "api.trakt.tv/shows/*/seasons?extended=episodes": "2d"

Might even work like offline mode now.

[glensc]

weird errors have appeared with cold cache

WARNING  Connection pool is full, discarding connection: 192-168-x-y.[redacted].plex.direct. Connection pool size: 10  
WARNING  Connection pool is full, discarding connection: 192-168-x-y.[redacted].plex.direct. Connection pool size: 10  
WARNING  Connection pool is full, discarding connection: 192-168-x-y.[redacted].plex.direct. Connection pool size: 10

comes from urllib3

site-packages/pip/_vendor/urllib3/connectionpool.py:                "Connection pool is full, discarding connection: %s. Connection pool size: %s",

EDIT: Asked question about it in requests-cache:

• stale_while_revalidate pool size and shutdown requests-cache/requests-cache#978

[glensc]

Heavy caching may result duplicate marking as watched, i.e plex state is cached, and now another Marking as watched would make duplicate watch.

Should re-fetch item before making actual changes.

[glensc]

Stale if error moved a to separate pr:

• Refactor: Enable stale_if_error, stale_while_revalidate #2132

[glensc]

It seems to be working locally for a while.

Also, as trakt reduced api calls per 5 minute limit, it might become useful to tune caching:

• Trakt API ratelimit using plextraktsync #2151