Add max_alt_alleles option to schema generation

[jeromekelleher]

Sometimes we do want to clip.

Probably the right thing to do here is to add a min_value and max_value values to the column schemas, and to np.clip() the values during sanitise.

[jeromekelleher]

Note that there are numerous thorny issue to deal with here, as we may not want to just keep the first k alleles but the k most frequent. Except for PL fields (for which we need local alleles anyway, #185) it doesn't see to cause too much problems, so I'm not sure this is much of a priority. Going to leave open for now, but we should probably close this issue after a while if nobody comes along with a compelling use-case.