A few metrics to consider

• distances between 2 keypoints (within each individual, e.g. dist_snout_rightear for each individual)
• distances between 2 individuals (across all pairs of the same keypoints, e.g. dist_individual1_individual2 contains snout-to-snout, rightear-to-rightear, tail-to-tail, etc. distances between individual1 and individual2)
• distances between different keypoints of different individuals (dist_individual1snout_individual2rear contains the distances between individual1's snout and individual2's rear)

See also

• SciPy's pdist, which returns a condensed distance matrix, and cdist
• relevant SO question on achieving pdist in xarray
• Matlab's pdist2