Fix a bug in VR jet overlap logic

There were a few things incorrect about the previous implementation here:

• The 'absolute' delta R was actaully the 'relative' delta R
• The 'relative' delta R was actually delta R / min_radius**2, i.e. garbage
• We should have been saving the absolute delta R for the pair that had the smallest relative delta R.

This should fix all that. Tagging @feickert who helped me understand the logic. Also tagging @miochoa.