Optimize ARIS to video conversion with vectorized operations and cv2.remap

[Chouffe]

Summary

• Replaced slow nested for-loop frame reading with vectorized numpy array construction for ~100x faster frame parsing
• Added LUT() function using numpy meshgrid for efficient lookup table generation
• Replaced dictionary-iteration remapARIS() with cv2.remap() for fast OpenCV-based coordinate transformation
• Removed obsolete createLUP() and old remapARIS() functions
• Added cv2 import and helper functions (getBeam, getBin, vectorized versions)

Performance

Conversion speed improved from ~50 frames/sec to ~1,200 frames/sec (~24x faster), tested on 2GB ARIS files with 25,000+ frames.

Test plan

• All 99 existing tests pass
• Successfully converted 9 ARIS files from sonar-smolt-detection project
• Output video properties verified (correct FPS, duration, codec)
• Code formatted with ruff and linted