Automatic Video Frame Resizing in edgeimpulsevideoinfer

[ferjm]

This PR enhances the edgeimpulsevideoinfer GStreamer element to automatically handle video frame resizing internally, eliminating the need for users to manually resize frames before inference.

Previously, users were required to ensure video frames matched the ML model's expected input dimensions (e.g., 224x224). This forced complex GStreamer pipelines with external videoscale elements and manual resolution management:

# Old approach - complex pipeline with manual resizing
gst-launch-1.0 qtiqmmfsrc camera=0 ! video/x-raw,format=NV12,width=1920,height=1080,framerate=30/1 ! \
  videoconvert ! videoscale ! video/x-raw,format=RGB,width=224,height=224 ! \
  edgeimpulsevideoinfer threshold="3.minScore=0.5" ! edgeimpulseoverlay ! \
  videoconvert ! videoscale ! video/x-raw,format=RGB,width=1920,height=1080 ! \
  waylandsink display="wayland-1" sync=false fullscreen=true

# New approach - clean, simple pipeline
gst-launch-1.0 qtiqmmfsrc camera=0 ! video/x-raw,format=NV12,width=1920,height=1080,framerate=30/1 ! \
  videoconvert ! edgeimpulsevideoinfer threshold="3.minScore=0.5" ! \
  edgeimpulseoverlay ! waylandsink display="wayland-1" sync=false fullscreen=true