feat(transcoder): rewrite on libav (node-av) bindings

[devchaudhary24k]

Summary

• Drop subprocess spawn("ffmpeg") in favor of in-process libav via node-av (MIT, prebuilt binaries — no system ffmpeg install required).
• Decode-once / fan-out ABR pipeline: single demuxer + decoder feeds an N-way split/asplit filter graph; one encoder + HLS muxer per rung. Kills 2-3× duplicate decode work and the h264 threaded-decoder races that flooded stderr.
• Graceful shutdown via new shared @vidcastx/queue/shutdown helper — SIGINT / SIGTERM / SIGHUP → worker.close(true) so tsx watch restarts release Redis locks cleanly.
• BullMQ lock tuning: lockDuration: 10m, lockRenewTime: 2m, stalledInterval: 60s — default 30s was flagging long transcodes as stalled.
• libav log filter (libav-log.ts) suppresses four known-harmless h264 thread-race patterns; everything else flows through with [libav] prefix.
• Poster fix: yuv420p,setparams=range=pc + color_range: "pc" to silence deprecated pixel-format warning.
• Hover preview fix: output-packet-count stop condition instead of trim filter (which EOF'd mid-pipeline and silently broke thumbnails).
• AAC tail drain: encoder.encodeAll(null) so the final 1024-sample partial frame ships instead of being dropped.
• Extended workers/transcoder/ROADMAP.md with HW-decoder design under §4 and new §8 for structured job logging; TODO summary at bottom.
• Unrelated chores bundled: db:studio --host=0.0.0.0, routeTree regen, IDEA.md rewrite as product-vision doc.

Test plan

• pnpm check-types — clean across all 11 workspaces
• pnpm --filter transcoder lint — clean
• Smoke: 10s testsrc clip → HLS + poster + preview produced, no reference picture missing spam in stderr
• Real BullMQ job end-to-end from apps/app upload → transcoder → ready state
• Observe progress ticks in BullMQ board (should be smooth per-frame, not 2-3 big jumps)
• Confirm MinIO layout: processed/<org>/<vid>/master.m3u8 + per-rung playlists/segments + thumbnails/ + previews/
• Video plays end-to-end in studio player
• Kill worker mid-job with Ctrl+C → confirm clean shutdown log + job returns to queue (no stalled lock)

Notes

• Target branch: main per request. Do not merge — leaving open for review.
• Roadmap items §1–§8 in workers/transcoder/ROADMAP.md track the remaining work beyond this PR.

[devchaudhary24k]

Review: parking this as a draft for now

I ran the libav pipeline against the FFmpeg CLI worker on dev, using test clips made with the FFmpeg binary that node-av ships. It type-checks and lints cleanly, but it isn't ready to replace the CLI worker yet.

Blockers

• Frame loss on any clip with audio. In a 10s clip, 190 of 300 frames came out (video 6.3s long, audio 8.5s). In a 90s clip, only 1536 of 2700 distinct frames. The cause: transcode.ts runs separate video and audio read loops on one demuxer, and node-av 5.2.3 silently drops a stream's packets once its queue reaches 100. libav-log.ts then hides the resulting h264 errors, even though they are real.
• Seeking uses a stream index instead of -1 (poster.ts, preview.ts, probe.ts). The offset is read as stream time units, so the poster and preview are taken near the end of the clip, and the per-title probe samples the wrong parts.
• No audio format conversion or downmix before AAC. PCM audio (common in .mov files) fails the whole job.
• Rotation metadata is ignored. Portrait phone video comes out sideways.
• Licensing. The prebuilt node-av.node contains a GPLv3 FFmpeg build with libx264 and runs inside the worker process. That needs a decision now that the repo is MIT.

Also

• The master playlist is written by hand, without CODECS, and BANDWIDTH leaves out audio.
• close(true) doesn't release the job lock. Combined with the 10-minute lock, a killed job stays stuck for about 10 minutes.
• probe.ts throws when the duration is unknown, which is common for WebM files recorded in a browser.
• There's an unvalidated as FFEncoderCodec cast, and errors are logged with only err.message.
• pnpm 11 allowBuilds covers only linux-x64, so the binding won't load on macOS or arm64.

Worth keeping: forcing 8-bit yuv420p output. The CLI worker on dev currently produces High 10 H.264 from 10-bit sources, which most browsers can't play. Also worth keeping: the shutdown helper idea and per-title encoding once the probe is correct. These are now tasks for the CLI worker in the V1 plan.

Fixing this branch is roughly 2–3 days of work. The branch stays as it is, so it can be picked up again as its own experiment after V1.