Add option to specify num_alleles when using solve_num_mismatches

[szhan]

Fix #385

[szhan]

I didn't update match_recombinant because it is marked for removal.

[szhan]

Also, in the tests, num_alleles is hard-coded to 5, because the functions are calling match_tsinfer with default option of deletions_as_missing, which is False.

[jeromekelleher]

let's default to n=5 in the signature to minimise breakage. It's really not going to matter much, as it's just very slightly altering the proba of choosing a recomb vs not.

[jeromekelleher]

I'm going to move the solving for mu/rho given num mismatches into match_tsfiner, so don't worry too much about details here.

[szhan]

I'm passing a specific value for num_alleles in match_samples because deletions_as_missing is specified.