fix(sabnzbd): keep failed jobs on the first history page

[firestaerter3]

Fixes #992

Sonarr and Radarr poll the SABnzbd history with start=0 and a fixed limit (DownloadClientHistoryLimit, default 60) and never read a later page. A tracked download that is in neither the queue nor that page is marked untrackable without a downloadFailed event, so the release is never blocklisted and never retried. During a burst of grabs, newer completed rows push failed rows off the page before the arr polls.

On start=0 the history handler now appends every failed row still kept in import_queue (same category and skip_arr_notification filters, capped at 1000) that the page does not already hold. The page itself, noofslots, later pages and the nzo_ids lookup are unchanged.

Tests:

• TestSABnzbdHistory_FailedRowOnFirstPageDuringBurst: one failed job older than 8 completed jobs, limit=5. Fails on the current code, passes with this change.
• Guard tests for later pages, no duplicate when the failed row is already on the page, and the nzo_ids lookup.
• TestSABnzbdFailedHistory_FiltersAndCap for the new repository query.

We have been running this on our instance since 2026-10-09. After deploying, page 0 for one category carried all 46 failed jobs out of 3541 history rows.