Adds an option to select the pcutoff for each bodypart individually for models trained with the PyTorch engine.

The pcutoff can be set as a float, a list or a dict. Examples for a project where there are 5 bodyparts: ["snout", "left_ear", "right_ear", "mid", "tail_base"].

Example 1: Using the same pcutoff for all bodyparts

deeplabcut.evaluate_network(config, ..., pcutoff=0.6),

Example 2: Using a list to set the pcutoff

deeplabcut.evaluate_network(config, ..., pcutoff=[0.6, 0.8, 0.8, 0.6, 0.6]),

Example 3: Changing the pcutoff for the ears to 0.8, and otherwise keeping the same values as set in the project config.yaml:

deeplabcut.evaluate_network(config, ..., pcutoff={"left_ear": 0.8, "right_ear": 0.8})

Tests were added for evaluation.