SIGN IN SIGN UP

Run targeted MLAS tests across CPU architectures (#33054)

### Description

- **CTest integration**
- Add opt-in registration for a focused MLAS FP16 and activation test
subset.
- **CI coverage**
  - Enable the subset once for x86, x64, and ARM64 PR CI.
  - Keep the full, long-running MLAS suite out of per-PR workflows.

### Motivation and Context

CI builds `onnxruntime_mlas_test` but never executes it, leaving MLAS
regressions undetected. This adds representative cross-architecture
coverage while keeping CI runtime reasonable.

<!-- START COPILOT CODING AGENT SUFFIX -->

- Fixes #33052

---------

Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
Co-authored-by: tianleiwu <30328909+tianleiwu@users.noreply.github.com>
C
Copilot committed
e1b899646599585d3d0d3efbe2af9bdaaa7ebc87
Parent: d32356b
Committed by GitHub <noreply@github.com> on 10/2/2026, 4:02:25 AM