[OSS CI] Integrate target determinator

[psiddh]

Objective
Evaluate and integrate target determinator into ExecuTorch's OSS CI so that only tests affected by a PR's changes are executed, reducing CI cost and latency.

Background
Target determinator is used by PyTorch core to intelligently select which tests to run based on the files changed in a PR. Instead of running the full test suite on every PR, it maps code changes to relevant test targets — significantly reducing CI time and compute cost without sacrificing signal quality.

Action Items

• Understand target determinator's current capabilities and integration requirements (talk to PyTorch CI infra team)

• Evaluate ExecuTorch's test-to-source mapping complexity — how well can TD work given our repo structure (buck + CMake + GH Actions)?

• Run a pilot: enable TD on a subset of workflows and measure:

  • Reduction in tests run per PR
  • CI time savings
  • Any missed regressions (false negatives)

• Define rollout plan: which workflows first, timeline, fallback strategy

• Full integration with monitoring for false-negative rate

Success Criteria

• Target determinator integrated on at least the top 3 most expensive workflows
• Measurable reduction in average CI time per PR (target: 30%+)
• False-negative rate < 1% (regressions not caught by selective testing)
• Clear rollout timeline documented

References
PyTorch Target Determinator: https://github.com/pytorch/pytorch/wiki/Testing-infrastructure-overview
ExecuTorch Cost Dashboard: https://hud.pytorch.org/cost_analysis?dateRange=7&granularity=day&groupby=workflow_name&chartType=stacked_bar&yAxis=cost&repos=pytorch%2Fexecutorch