[Kimlimp]

Summary

Eliminate unnecessary mark-and-sweep phases on a genuine first M3U import into an empty store, while preserving interrupted-import retry recovery and normal refresh pruning.

Root Cause

During a cold M3U import into an empty store, every row written was freshly inserted. Running 5 full mark-and-sweep phases (pruning live streams, movies, series, episodes, categories) walked every newly inserted record against seen ID sets only to delete 0 rows, consuming significant time and memory holding seen-sets until the end of the last sweep.

Changes

1. Pre-import emptiness detection (ContentSyncManager+Prune.swift): Added playlistHasExistingCatalogRows(playlistId:) with bounded fetchLimit: 1 across LiveStream, Movie, Series, and Episode. This avoids relying on playlist.lastSyncDate, which is nil even after an interrupted/partial import.
2. State propagation (ContentSyncManager+M3UImportState.swift & +M3UFields.swift): Seeded state.hadExistingContent = !state.knownCategories.isEmpty || playlistHasExistingCatalogRows(playlistId: playlistId) before writes begin.
3. Prune bypass (ContentSyncManager+M3UPrune.swift): Guarded pruneStaleM3URows(playlistId:state:) on state.hadExistingContent. When false, immediately clears seen-sets and returns without running sweeps.
4. Tests (ContentSyncPruneTests.swift & M3USyncTests.swift):
   • Unit tests verifying emptiness check isolation, skip behavior, and reimport pruning.
   • Integration test verifying that an interrupted import with committed rows followed by a shorter retry properly executes sweeps and prunes stale rows.
   • Benchmark test testReimportProviderShapedPlaylist in M3UColdImportBenchmarks.swift.

Validation

• ContentSyncPruneTests: 5/5 passed (0 failures).
• M3USyncTests: 12/12 passed (0 failures).
• 50k fixture benchmark: monotonic clock reduced from 14.175 s to 12.671 s (-1.504 s / ~11% reduction), with all prune signpost phases completely skipped.