Fix plex.tv DNS blips blocking array restore, clarify locked admin toggles

[Brandon-Haney]

Fixes #197

What's happening

A user reported this during a scheduled run:

ERROR - [PLEX API] Error (load user tokens): ... Failed to resolve 'plex.tv' ([Errno -5] No address associated with hostname)
WARNING - [PLEX API] plex.tv unreachable - using cached user data only
WARNING - [PLEX API] Watchlist data will be incomplete - array restore will be skipped

Two problems fell out of it, plus a UI gap.

1. The first plex.tv call of a run had no retry

_get_main_account() called myPlexAccount() bare. It was the only plex.tv call in core/plex_api.py not going through _retry_plextv_call, which already retries requests.ConnectionError three times with 2s/4s backoff. A NameResolutionError surfaces as exactly that exception, so a brief DNS hiccup took out the whole run when a retry would have gone straight through.

Now wrapped in the existing helper. Non-transient errors (auth, parse) still fail fast on the first attempt.

2. A plex.tv outage blocked array restore even with watchlist disabled

The failure path set _watchlist_data_complete = False unconditionally, and _process_media() uses that flag to skip moving watched files back to the array. With watchlist caching turned off, no cached file is held by the watchlist, so there was nothing incomplete to guard against. Users who don't use the feature at all were still losing their array restore whenever plex.tv had a moment.

PlexManager now takes watchlist_enabled and only flags watchlist data when watchlist caching is actually on. _process_media() checks watchlist_toggle alongside the flag as a second gate. The OnDeck guard is unchanged, and behaviour with watchlist enabled is identical to before.

3. The locked admin toggles didn't explain themselves

The admin row's OnDeck and Watchlist toggles are disabled by design, but nothing said why or where to go instead, so the reporter thought they'd hit a bug. Added a tooltip on each, and a hint under the table linking to the global switch at Settings > Cache > Enable Watchlist Caching, which is what actually turns watchlist caching off for everyone.

Testing

New tests/test_plextv_main_account_resilience.py, 7 tests:

• transient DNS failure then success returns the account and leaves both health flags clean
• sustained failure stops at PLEXTV_MAX_RETRIES and reports plex.tv unreachable
• happy path makes exactly one call
• auth-shaped errors aren't retried
• watchlist enabled still flags data incomplete
• watchlist disabled leaves data complete so restore proceeds
• constructor defaults watchlist_enabled=True for existing callers

Full suite: 1207 passed.

Manual verification

1. Settings > Users, confirm the admin row's toggles show a tooltip on hover and the hint below links to the cache tab.
2. Click the hint link, confirm it lands on Enable Watchlist Caching and flashes it.
3. Turn off Enable Watchlist Caching, block DNS for plex.tv, run a cache operation. The log should show "plex.tv unreachable" but no "array restore will be skipped", and watched files should still move back to the array.
4. Turn Enable Watchlist Caching back on and repeat. The skip warning should return, matching current behaviour.