v1.18.0 — NFO titles your media server can actually match

Everything here came out of the Dispatcharr Discord thread. No breaking changes; nothing needs re-generating.

• Provider tags are stripped from NFO titles. Titles like 4K-A+ Blade Runner 2049, |EN| The Matrix or [4K] Dune (2025) were written into <title> verbatim, and since Jellyfin/Emby treat an NFO title as authoritative, the item stayed mis-titled no matter what TMDB said. Leading provider tags (A+, 4K-A+, |EN|, [4K]), edge quality tokens and trailing years are now removed.

  Every step is anchored to the start or end of the title, deliberately gentler than the folder-name cleanup, and the result was checked against a live 6,877-title catalogue: 0 titles emptied, and the only non-year changes were correct ones. Names that look like junk but aren't all survive — Love+War, Genera+ion, NTSF:SD:SUV::, WWII in HD, Chicago P.D., Marvel's Agents of S.H.I.E.L.D., [REC] 2, Blade Runner 2049, Bali 2002.

• New Omit <title> from NFO files setting (off by default). If your provider's naming is too mangled for any cleanup to rescue, turn this on: movie and tvshow NFOs are written without a <title> element at all, so Jellyfin/Emby fall back to matching on the folder name and take the title from TMDB. Episode NFOs always keep their title — an episode has nothing else to identify it by.

• Duplicate episodes are collapsed (reported by rammboslice). When the same episode is reachable through more than one M3U relation, the plugin was writing a .strm per relation; with Omit stream_id from URLs on, those files resolve to identical URLs — so you'd get S01E05, S01E05 (2), and so on, all playing the same thing. Episodes are now deduplicated by episode UUID before writing, keeping the first relation in a stable order, and the run log says collapsed N duplicate episode relation(s) when it happens.

• Bigger series batch sizes — 50, 100 and 250 join the existing options, for people running large libraries who don't want to click Generate a dozen times. The help text now warns that anything past ~25 risks the UI 504 (the job keeps running server-side regardless) and points at the schedule as the better answer for a full library.

• Missing TMDB IDs are now reported. With Append TMDB ID to folder names on, items lacking a TMDB ID in Dispatcharr were silently written without a tag, which looks like the setting is broken. The summary now counts them: N item(s) had no TMDB ID.

23 new unit tests (224 total, was 201).