Fix DTS repair escalation and idet parse regressions

[ptr727]

Follow-ups from the 3.20→3.21 regression diff. Two files-flip-to-failed regressions plus a diagnostics/logging improvement.

1. idet parse robustness (interlace detection)

VC1 - National Lampoon's Christmas Vacation regressed to an Error (Failed to parse idet output → Failed to count interlaced frames), aborting the file before it could be re-encoded.

Root cause (confirmed against ground-truth ffmpeg output): the parser required the three idet stat lines contiguous. The idet command uses InputOptions.Default(), which adds -fflags +genpts; on a source with non-monotonic DTS the -f null muxer emits "non monotonically increasing dts to muxer" warnings interleaved with the idet stats, and a full-file scan decodes enough packets that a warning lands between the stat lines → matches.Count == 0.

Fix: match each stat line independently and take the last of each (robust to interleaving; still selects the final counts over idet's early all-zero pass). Log the raw idet output on a parse failure. An idet failure stays a hard error that aborts the file — it is unexpected, so it should surface as a bug, not be masked. Regression test added with warnings interspersed between the stat lines.

2. DTS repair escalation

Six files that were Repaired, Verified pre-3.21 became RepairFailed (50 First Dates, Kill Bill, Bitrate Love Island, Running Wild, Diplo, 90 Day). A non-monotonic DTS that the lossless audio setts repair couldn't fix (video-stream or post-decode) was marked RepairFailed with no further attempt; a full re-encode fixes these and the pre-3.21 pipeline relied on it.

Fix: the DTS path now escalates through the standard tiers — surgical lossless setts → plain remux → re-encode — marking Repaired on the first tier whose re-verify is clean, RepairFailed only when all fail. Added the missing remux tier to RepairAndReVerify so the DTS and decode-error paths share one surgical → remux → re-encode ladder.

3. Per-file processing time

Time each file's task in the driver and log it on the After line (elapsed before the file name), sharing one HH:mm:ss.fff formatter with the run summary (hours from TotalHours, no multi-day wrap).

Validation

Local build + develop image ffmpeg 8.0.1, --parallel --testsnippets:

• VC1 clip → ReMuxed, ReEncoded, Verified, no idet failure (was abort).
• Diplo (post-decode DTS) → Repaired, Verified via setts→remux→re-encode (was RepairFailed).

217 tests pass; build + format clean.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR addresses two media-processing regressions introduced in the 3.20 -> 3.21 transition (idet parsing and DTS repair escalation) and adds more consistent per-file timing diagnostics in the processing driver.

Changes:

• Make ffmpeg idet parsing robust to interleaved stderr output by matching each stats line independently, and log the raw output on parse failure.
• Escalate non-monotonic DTS repairs through the standard repair tiers (lossless setts -> remux -> re-encode) instead of stopping after setts failure.
• Add per-file elapsed timing (plus consistent duration formatting) to driver logging, with targeted tests.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
PlexCleaner/FfMpegIdetInfo.cs	Reworks idet parsing to tolerate interleaved warnings and improves failure diagnostics.
PlexCleaner/ProcessFile.cs	Adds a remux repair tier and escalates DTS repair through remux/re-encode fallbacks.
PlexCleaner/ProcessDriver.cs	Times per-file processing and introduces a shared FormatDuration() for consistent elapsed logging.
PlexCleanerTests/FfMpegIdetParsingTests.cs	Adds regression coverage for idet output with interleaved non-monotonic DTS warnings and updates parsing assertions.
PlexCleanerTests/ProcessDriverTests.cs	Adds unit coverage for FormatDuration() millisecond precision and multi-day hour accumulation.