Improve vignette variable naming, patterns, dependencies on data-raw

[brookslogan]

Debugging vignette errors is a made more difficult by:

• non-explanatory variable naming
• eval=FALSE chunks that have fallen out of sync with the actual data being extracted from objects in data-raw
• bindings being overwritten, so repeating error behavior sometimes repeating the initial binding + some rebindings while avoiding eval=FALSE chunks

Improve vignette variable naming while reducing the amount of binding&object mutation that occurs. Add a development guide and/or notes in data-raw that instruct developers to update vignettes accordingly whenever updating data-raw scripts.

Also consider turning echo=FALSE chunks loading data objects into visible chunks with an explanation for the reader. (So they can run things more quickly.)

[brookslogan]

Or, for the nonevaluated API + nonechoed data object chunk pairs, consider just eval+echo the chunk using the example data.

[dshemetov]

Going to close this in favor of #363