`parse_tax_data`: incorperate rank info

[zachary-foster]

Currently, the Taxon objects created for the taxmap output of parse_tax_data do not add rank information, although it is often available in various forms. This means taxon_ranks() does not work as expected (see grunwaldlab/metacoder#188 and grunwaldlab/metacoder#189).

The challenge here is the diversity of inputs to parse_tax_data and the different ways ranks can be encoded:

• A character vector of classifications (e.g. Animalia;Chordata;Mammalia;Primates;Hominidae) might have the rank information embedded (e.g. k_Animalia;p_Chordata;c_Mammalia;o_Primates;c_Hominidae), so a new class_key option like "taxon_rank" could be added.
• A data.frame could have a column like the character vector above, or could have one column per rank, which would mean the ranks are the column names. So either a new class_key option or a new option that indicates ranks should be taken from column names
• A list of data.frames, one per classification, would store its rank info in a column, so the name of the column with rank information would need to be passed to a new option.
• A list of characters, one per classification, could store its rank info as the names of the vectors.

This means there are 4 ways to encode rank and handling all would require at least 2 new options and a new class_key value. The class_key value is intuitive and would not clutter the help page any, but I hesitate to add 2-3 options to handle rank.

Currently, I am thinking of doing the following:

• Add a value to class_key called "taxon_rank" for when the rank and taxon name are together in a vector/column (quite common).
• Add a TRUE/FALSE option called named_by_rank so users can say when column/vector names are ranks. ( somewhat common)

This will not handle lists of data.frames, but that is the least common input type. Maybe another option called "rank_col" could be added in the future.