Document and/or improve protein MinHash API.

[ctb]

As seen in #701, the way the MinHash object interface works for protein MinHashes is undocumented and quite possibly just stooopid.

A few issues:

• per this comment, MinHash.add_sequence assumes the input is DNA and does a translation of the provided sequence before hashing it, rather than adding the protein k-mers directly. The current "correct" code to use is MinHash.add_protein which assumes the input is amino acids. This should be documented / demonstrated in the API documentation, doc/api-example.md.
• related, perhaps add_sequence should be named add_dna_sequence...
• separately, per this comment, MinHash.add_protein divides the k-mer size by 3 before generating the k-mers. This is also quite hidden behavior and at the least needs to be documented. Perhaps we should fix this somehow.
• in the provided example in Setting "scaled=1" doesn't retrieve all protein k-mers #701, I'm not sure why add_sequence doesn't complain when it's given a bunch of non-ACTG characters. That should also be looked into.

[ctb]

ref #885 maybe

[ctb]

Bunch of stuff in this comment to pay attention to - #999 (comment)