Add support for ``individuals`` to divergence_matrix

[jeromekelleher]

Currently the divergence matrix supports a list of samples. It would also be useful to support individuals as a mutually exclusive option. Initially we can implement this by post-processing the matrix we get back from the corresponding samples, but we'll probably want to do it natively in C at some point to save memory.

See https://github.com/tskit-dev/tskit/pull/2736#issuecomment-1535099324