Support for action queries in Eva

[pchunduri6]

1. Added GROUP BY operator for video inputs - GROUP BY "16f" stacks 16 frames into a segment.
2. Added Aggregate Expressions:
   • FIRST - Return the first element in the group
   • LAST - Return the last element in the group
   • SEGMENT - Stack 3d frame inputs into 4d.
3. Added the MVIT (multi-vision transformer-based) ActionRecognition UDF with SOTA action classification accuracy.

Queries supported are of the form:

SELECT FIRST(id), MVITActionRecognition(SEGMENT(data)) FROM Actions GROUP BY '16f';

[LordDarkula]

Left a couple comments. Good for the most part.

Another approved review.