Convert gvapython samples to standalone Python GStreamer elements usi…

[jmotow]

Summary

Converts two gvapython-based samples into standalone GStreamer Python elements using the GstAnalytics metadata API, as a step toward removing gvapython element support from the DLS project.

Changes

New samples in samples/gstreamer/gst_launch/python-elements/

save_frames_with_ROI_only

• gvaframesaver_py — custom BaseTransform element that reads GstAnalytics.ODMtd detection metadata and saves annotated frames to disk
• Replaces: gvapython module=... class=FrameSaver function=process_frame

face_detection_and_classification

• gvaagelogger_py — custom BaseTransform element that reads GstAnalytics.ClsMtd classification metadata and logs age values to file
• Replaces: 3× gvapython elements (SSD post-proc, age/gender post-proc, age logger) with gvadetect + gvaclassify + gvaagelogger_py

Key design decisions

• Follows the same plugin pattern as samples/gstreamer/python/smart_nvr (custom elements in plugins/python/)
• Uses GstAnalytics API (ODMtd, ClsMtd) instead of gstgva.VideoFrame API
• Shell scripts auto-add local plugins/ to GST_PLUGIN_PATH — no external setup needed
• Element properties are configurable via gst-launch-1.0 syntax

Other changes

• Updated samples/gstreamer/README.md with new sample entries
• Fixed setenv.sh to resolve script directory via dirname instead of $(pwd)

Testing

• face_detection_and_classification.sh — ~212 FPS on CPU, 3410+ age entries logged to /tmp/age_log.txt
• save_frames_with_roi.sh — frames with bounding boxes saved to saved_frames/
• pylint 10.00/10 on both Python files (with project pylintrc)