add clarity to document and tests to enforce categorical datatypes with string/object values.

[ejmolinelli]

Motivation

There is some ambiguity in the data specification and corresponding test suite regarding the string-encoded categorical datatypes; i.e. the column is of type CategoricalDtype(..., categories_dtype=string) or CategoricalDtype(..., categories_dtype=object).

Without clarity, idiosyncrasies and edge cases exist both in the internal library and with third party integration testers (i.e. lattice team). See this PR #1142 for an example of an issue that emerged.

Additional Details

In the specification for 5.3 assay_ontology_term_id field, it reads:

categorical with `str` categories. This MUST be an EFO term and either:

... meanwhile, for the {column}_colors field, it reads:

{column}_colors where {column} MUST be the name of a `category` data type column in obs that
is annotated by the data submitter or curator.

Definition of Done

All dataframe columns in the schema specification document that are expected to be categorical are clearly identified as such. Test fixtures in the testing suite should be updated accordingly and tests should be introduced to test for categorical vs string encoded columns.

Tasks