perf(dgw): improve real-time performance of session shadowing

[irvingoujAtDevolution]

Fixes session shadowing playback stalls that occur when the video re-encoding pipeline falls behind real-time.

The main fix is an encoding strategy change: previously every re-encoded frame was force-keyframed, which is extremely CPU-intensive. Now only the first frame after the cut point is force-keyframed, and subsequent frames use normal inter-frame prediction, dramatically reducing encoder workload. Pre-cut-point frames are now decode-only instead of running the full encode pipeline and discarding the output. Encode duration is also derived from actual frame timestamps instead of a hardcoded value.

On top of this, the streamer adaptively skips encoding when falling behind real-time as a safety net, and the shadow player correctly seeks to the live edge using buffered ranges instead of seekable ranges.

Issue: DGW-341

[github-actions]

Let maintainers know that an action is required on their side

• Add the label release-required Please cut a new release (Devolutions Gateway, Devolutions Agent, Jetsocat, PowerShell module) when you request a maintainer to cut a new release (Devolutions Gateway, Devolutions Agent, Jetsocat, PowerShell module)

• Add the label release-blocker Follow-up is required before cutting a new release if a follow-up is required before cutting a new release

• Add the label publish-required Please publish libraries (`Devolutions.Gateway.Utils`, OpenAPI clients, etc) when you request a maintainer to publish libraries (Devolutions.Gateway.Utils, OpenAPI clients, etc.)

• Add the label publish-blocker Follow-up is required before publishing libraries if a follow-up is required before publishing libraries

[Copilot]

Pull request overview

This PR adds diagnostics, tests, and benchmarks around the WebM re-encoding pipeline to reproduce and measure the “progress stall” symptom and validate timeline correctness for live session playback.

Changes:

• Introduces end-to-end correctness and performance tests for webm_stream, including live-growing input simulation and VPX-only reproductions of keyframe-related stalls.
• Adds gated perf-diagnostics logging and statistics around the re-encoding and streaming path (VPX encoder/decoder, tag handling, and channel writer).
• Provides a bench_support module and a Criterion benchmark (vpx_reencode) plus documentation and tooling (README, Justfile, Cargo features) for running targeted performance measurements.

Reviewed changes

Copilot reviewed 16 out of 18 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
crates/video-streamer/tests/webm_stream_perf.rs	Adds an ignored perf test that simulates attach-at-20s on a growing WebM, writes output to disk, and asserts a low media-vs-wall-time ratio to reproduce the “progress stall” behavior.
crates/video-streamer/tests/webm_stream_correctness.rs	Adds a suite of ignored async tests to validate timeline reset to 0, monotonic/simpleblock progress, metadata behavior on Start/Pull, EOF pause/resume handling, slow client behavior, and client disconnect semantics for webm_stream.
crates/video-streamer/tests/vpx_kf_spike_repro_first_kf_only.rs	Adds a VPX-level control experiment that force-keyframes only the first post-cut frame to compare encode performance vs. the full forced-keyframe scenario.
crates/video-streamer/tests/vpx_kf_spike_repro.rs	Adds a minimal VPX-only reproduction of the progress stall by continuously force-keyframing frames after a cut at ~20s.
crates/video-streamer/tests/support/mod.rs	Introduces shared test infrastructure: in-memory websocket, live file writer, stream harness for webm_stream, global semaphore, XMF initialization, and WebM timestamp extraction helpers.
crates/video-streamer/tests/common/vpx_kf_spike.rs	Adds common helpers for VPX tests to parse stream config and collect per-frame timestamps/keyframe flags for cut-point calculations.
crates/video-streamer/tests/common/mod.rs	Re-exports the vpx_kf_spike helpers for use in VPX-related tests.
crates/video-streamer/src/streamer/tag_writers.rs	Adds perf-logging around VPX encoder/decoder construction and re-encoding, tracks per-frame timing and stream realtime ratio behind a perf-diagnostics feature, and enriches error handling (including channel-closed detection) in CutClusterWriter.
crates/video-streamer/src/streamer/mod.rs	Switches some debug!/trace! logging to perf_debug!/perf_trace! for header logging and EOF retry loops, tying them into the new diagnostics feature gate.
crates/video-streamer/src/streamer/iter.rs	Adds perf_trace!/perf_debug! logging for keyframe discovery and rollback operations to aid debugging cut-point selection.
crates/video-streamer/src/streamer/channel_writer.rs	Extends ChannelWriter and ChannelWriterReceiver with optional atomic counters and perf logging for writes/reads and total bytes, plus more informative diagnostics on channel closure.
crates/video-streamer/src/lib.rs	Defines perf_trace! and perf_debug! macros gated by the perf-diagnostics feature, re-exports core streaming types, and exposes the new bench_support module under a bench feature.
crates/video-streamer/src/bench_support.rs	Adds reusable helpers to re-encode the first N tags of a WebM stream (or until a wall-clock deadline), returning stats on tags, frames, media span, and bytes written for use in benchmarks.
crates/video-streamer/benches/vpx_reencode.rs	Introduces a Criterion benchmark that repeatedly re-encodes the first 500 tags of a test WebM, with optional verbose stdout logging and per-iteration timeouts to bound work.
crates/video-streamer/README.md	Documents prerequisites (XMF, environment vars), how to run correctness tests and ignored tests, how to enable perf-diagnostics, and how to run the new benchmarks with bench/verbose options.
crates/video-streamer/Cargo.toml	Adds perf-diagnostics and bench features, the criterion dev-dependency, and registers the vpx_reencode bench target without a harness.
Justfile	Adds a Windows/PowerShell-oriented test recipe (plus test-streamer alias) to run video-streamer correctness tests with logging and an isolated target directory.
Cargo.lock	Updates the lockfile to include Criterion and its transitive dependencies, as well as other version bumps consistent with the new dev tooling.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[CBenoit]

Approving so you can merge on your end

[CBenoit]

@irvingoujAtDevolution Before merging, update the PR into something more proper to be in the changelog (e.g.: perf(dgw): improve performance of session shadowing).