QVAC-17095 feat: add Whisper performance benchmark workflows

[ogad-tether]

Note: be concise and prefer bullet points.

🎯 What problem does this PR solve?

• Adds a Whisper desktop RTF benchmarking path across the supported CI runner matrix so performance findings can be collected per backend and runner.
• Ensures manual benchmark and benchmark-workflow runs can exercise the branch addon/prebuilds instead of silently falling back to the published package.

📝 How does it solve it?

• Adds a desktop Whisper RTF benchmark harness, matrix runner, and aggregation tooling that emit structured JSON artifacts for cross-runner comparison.
• Updates the desktop Whisper workflow and adds a dedicated benchmark-orchestration workflow to run prebuilds, benchmark-only desktop E2E paths, and publish one consolidated findings report.

🧪 How was it tested?

• npx standard "packages/qvac-lib-infer-whispercpp/test/benchmark/rtf-benchmark.test.js" "packages/qvac-lib-infer-whispercpp/scripts/run-rtf-benchmark-matrix.js" "scripts/perf-report/aggregate-whisper-rtf.js"
• node --check packages/qvac-lib-infer-whispercpp/test/benchmark/rtf-benchmark.test.js
• node --check packages/qvac-lib-infer-whispercpp/scripts/run-rtf-benchmark-matrix.js
• node --check scripts/perf-report/aggregate-whisper-rtf.js
• ruby -e "require 'yaml'; YAML.load_file('.github/workflows/integration-test-qvac-lib-infer-whispercpp.yml'); YAML.load_file('.github/workflows/benchmark-performance-qvac-lib-infer-whispercpp.yml'); YAML.load_file('.github/workflows/benchmark-qvac-lib-infer-whispercpp.yml'); puts 'ok'"
• Successful desktop benchmark run: https://github.com/tetherto/qvac/actions/runs/24777285232
• Full upstream verification rerun on reduced branch: https://github.com/tetherto/qvac/actions/runs/24785460222/

📊 Desktop Benchmark Results

Desktop results below come from the successful benchmark run:

• https://github.com/tetherto/qvac/actions/runs/24777285232

Device	Platform	Model	GPU	Backend	Mean RTF	P50	P95	Mean Wall (ms)	Tokens/s
ai-run-linux-gpu	linux-x64	ggml-tiny	cpu	cpu	0.0648	0.0647	0.0651	1307	52.1
ai-run-linux-gpu	linux-x64	ggml-tiny	gpu	cuda	0.0647	0.0647	0.0648	1305	52.2
macos-14-xlarge	darwin-arm64	ggml-tiny	cpu	cpu	0.0254	0.0252	0.0261	512	133.2
macos-14-xlarge	darwin-arm64	ggml-tiny	gpu	coreml	0.1001	0.1000	0.1020	2016	33.8
macos-15-large	darwin-x64	ggml-tiny	cpu	cpu	0.0639	0.0636	0.0674	1288	53.0
macos-15-large	darwin-x64	ggml-tiny	gpu	coreml	1.2256	1.1552	1.3663	24679	2.8
ubuntu-22.04	linux-x64	ggml-tiny	cpu	cpu	0.0521	0.0472	0.0677	1051	66.9
ubuntu-24.04-arm	linux-arm64	ggml-tiny	cpu	cpu	0.1392	0.1392	0.1392	2803	24.3
windows-2022	win32-x64	ggml-tiny	cpu	cpu	0.0483	0.0485	0.0498	975	69.9
windows-2022	win32-x64	ggml-tiny	gpu	directml	0.0476	0.0463	0.0511	960	71.2

[GustavoA1604]

/review

[github-actions]

Tier-based Approval Status

**PR Tier:** TIER1

**Current Status:** ✅ APPROVED

**Requirements:**
- 1 Team Member approval ✅ (1/1)
- 1 Team Lead OR Management approval ✅ (1/1)



---
*This comment is automatically updated when reviews change.*