Fix collect_data; route kind:benchmarks sections

[dcvijeticTT]

Update benchmark collection to support tt-inference-server’s report format introduced at the end of July, while preserving legacy reports.

• Route kind: "benchmarks" sections through the detailed mapper, preserving metrics, operating points, tool metadata, and target checks. Now kind "benchmarks" gets tagged with run_type: benchmarks in the DB
• Add missing image/audio metrics and expand Whisper sweep records.
• Filter report filenames and improve handling of unavailable or invalid metric values.

[Copilot]

🟢 Approval recommended

No unresolved review issues were identified.

Pull request overview

Updates collect_data to support kind: "benchmarks" sections, broader report formats, and partial-failure handling.

Changes:

• Filters benchmark report filenames.
• Expands metric mapping and validation.
• Routes structured benchmark data through the detailed mapper.
• Tracks failures without blocking report uploads.

File summaries

File	Summary
.github/actions/collect_data/test/test_benchmark_mapper.py	Adds coverage for routing, metrics, discovery, and failure handling.
.github/actions/collect_data/src/shared.py	Adds failure-state reset support.
.github/actions/collect_data/src/generate_data.py	Resets and reports partial collection failures.
.github/actions/collect_data/src/benchmark.py	Implements report filtering, routing, mapping, and validation.

Review details

• Files reviewed: 4/5 changed files
• Comments generated: 0
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

🔵 Needs a closer look

Unresolved moderate issues affect report determinism and migration correctness.

Review details

Suppressed comments (1)

.github/actions/collect_data/src/benchmark.py:70

• When both model_spec_*.json and runtime_model_spec_*.json exist for one job, this whitelist puts both paths into the same report group, but create_json_from_report selects whichever path happens to come first and parses the other as a benchmark report. Because os.walk does not guarantee that order, the model identity/config can vary between runs; select a canonical spec deterministically and exclude all other spec files from report processing.

    "model_spec_",
    "runtime_model_spec_",

• Files reviewed: 4/5 changed files
• Comments generated: 0 new
• Review effort level: Lite

[Copilot]

🟡 Changes recommended

Deterministically select one model-spec file and skip all model-spec files during benchmark report iteration.

Get a fresh assessment by requesting another Copilot review.

Review details

• Files reviewed: 4/5 changed files
• Comments generated: 1
• Review effort level: Lite

[vmilosevic]

Please fill in the PR description with what is changed, why and how, few sentences is enough, thanks!