Fix peg-insert-side-v3-policy.

[raldone01]

Should also fix the ci since for some reason the peg-insert-side-v3 policy is (sometimes?) worse than 80%.

The improved policy is more consistent and quicker too!

Current New Policy (ORANGE) Success Rate: 100.00%, Old Policy (BLUE) Success Rate: 82.40% after 1000 episodes
Current Average Step to Success: New Policy (ORANGE): 77.903, Old Policy (BLUE): 106.47572815533981 after 1000 episodes
Current Min Step to Success: New Policy (ORANGE): 48, Old Policy (BLUE): 51 after 1000 episodes
Current Max Step to Success: New Policy (ORANGE): 106, Old Policy (BLUE): 491 after 1000 episodes

Here are a few old failing cases:

peg-insert-1.mp4

peg-insert-2.mp4

peg-insert-3.mp4

[reginald-mclean]

@raldone01 I think I finally tracked down the real issue. mujoco>=3.5.0 has breaking changes where collisions calculations were changed.

[jkterry1]

@raldone01 can you please get CI to pass? At a minimum precommit is failing