fix: skip Trakt lookups (Methods 2/3) when TRAKT_CLIENT_ID is unset

[kilo-WATT]

Bug

process_media_item() has 4 methods for resolving an item to a TMDB ID:

1. Direct TMDB ID (from the source list)
2. IMDB ID → Trakt → TMDB ID
3. Title/Year → Trakt → TMDB ID
4. Overseerr's own title search (fallback) — referring to ListSync's own naming here (OVERSEERR_URL/OVERSEERR_API_KEY, overseerr_client); my own testing was actually against Seerr (the Overseerr-API-compatible successor project), not Overseerr itself

Methods 2 and 3 call the Trakt API unconditionally — even when TRAKT_CLIENT_ID is completely unset. Since Trakt is meant to be optional (Method 4 exists specifically as the no-Trakt fallback), this means every item still attempts a live call to api.trakt.tv that's guaranteed to fail before falling through.

This isn't just a fast failure. It can trip Trakt's own rate limiter, which triggers a 300-second sleep per item (see the 429 handling already in providers/trakt.py). Running a sync against a few hundred items with no Trakt configured can turn what should be a couple-minute sync into one that effectively never completes.

Repro

• Fresh install, TRAKT_CLIENT_ID left blank (only MDBLIST_LISTS configured)
• Ran a sync against a ~540 item MDBList list
• Log filled with repeated:

  ⚠️  Trakt API rate limit hit. Waiting 300 seconds...
  

• Sync did not meaningfully progress until Trakt calls were manually disabled

This is likely related to what's being reported in #85 and #68 — Trakt has been increasingly hostile to unauthenticated/misconfigured callers, so hitting it unnecessarily is more costly now than it used to be.

Fix

• Added is_trakt_configured() to providers/trakt.py — checks config/env for a Trakt Client ID without raising.
• Gated Methods 2 and 3 on it in main.py. When Trakt isn't configured, each logs a clear skip message and falls straight through (Method 2 → Method 3 → Method 4) instead of attempting the network call.
• No change in behavior when Trakt is configured — same 4-method chain as before.

Testing

• python3 -m py_compile clean on both changed files
• Verified in a real deployment (Seerr as the request backend): same ~540 item list, same missing TRAKT_CLIENT_ID, sync now completes normally instead of stalling on rate-limit sleeps