Switch from `DRY_RUN` to `--fix`

[tarkatronic]

In keeping with the thought path of this being a linter we should, by default, only report out on errors. That somewhat obviates the mechanism of a "dry run", and instead changes us to the more explicit operation of a "fix", such as with eslint --fix.

This will probably have to rely on first implementing an actual CLI library such as oclif, yargs or commander. I don't know the pros/cons of each, so happy to take recommendations.