Updating CVML CI tests to reflect instruction updates in PR #236

[sreeram-11]

1. Enabling Linux support
2. Updated opencv.md by removing steps 5–7 (no longer setting permanent env variables like OPENCV_INSTALL_ROOT, OpenCV_DIR, or modifying system library paths)
3. Updated the tests based on the PR Update CVML playbook to match official SDK documentation #236
   • Remove Git/Git LFS/clone checks because the updated README no longer clones RyzenAI-SW; it downloads and extracts the Ryzen AI CVML Library ZIP and sets AMD_CVML_SDK_ROOT.
   • Do not require NPU driver in the prereq test. Just print whether it is present. The updated README says CVML should fall back to GPU if NPU driver is not installed.
   • Do not set/check OpenCV_DIR. The updated README only uses OPENCV_INSTALL_ROOT in the build command, so CI should not require OpenCV_DIR.
   • Use AMD_CVML_SDK_ROOT as the CVML root, not a cloned repo. The updated README tells users to extract to C:\RyzenAI-Library or ~/RyzenAI-Library and set AMD_CVML_SDK_ROOT.
4. Removed opencv.md based on the comment: https://github.com/amd/playbooks/pull/236/changes#r3311358870

IMPORTANT note on Linux Runner Setup

• Initially, the same cvml-build-and-face-detection-linux test was passing from manul terminal run but failing from GitHub Actions workflow.
• The GitHub Actions runner service has a lower locked-memory / mmap-related resource limit than the interactive terminal.
• The Vitis AI NPU backend tries to allocate/map a 64 MB GE buffer, fails with EAGAIN / Resource temporarily unavailable, and then the native runtime segfaults.
• Solution: Set LimitMEMLOCK=infinity for the actions runner service.
  1. Run sudo systemctl edit actions.runner.<SERVICE-NAME>.service and add the lines:
  [Service]
LimitMEMLOCK=infinity
LimitNOFILE=1048576
  2. sudo systemctl daemon-reload
  3. sudo systemctl restart actions.runner.<SERVICE-NAME>.service

[danielholanda]

Awesome!