Consider how we want to handle display strings for codings

[mikix]

Currently, in an effort to be graceful around janky data sources, data-metrics entirely ignores display values - in the few cases (race & ethnicity) it renders a human readable string from a code, it hardcodes them.

Otherwise, it just shows the code value.

This is fine. But could be nicer. Do we want to try to read display? Do we want to pull in a system vocab/dictionary to provide display values? Do we want to leave that for the display layer (in which case, we should probably push along a tuple of code & system)?

[dogversioning]

if we have the UMLS study in place, i believe you can use that for code lookup, though it does introduce a large dependency (and one that may be hard to run locally).