By having all phylogenetic workflows start from two lists of inputs
(`config.inputs`, `config.additional_inputs`) we enable a broad range of
uses with a consistent interface.

1. Using local ingest files is trivial (see added docs) and doesn't need
   a bunch of special-cased logic that is prone to falling out of date
   (as it had indeed done)
2. Adding extra / private data follows the similar pattern, with an
   additional config list being used so that we are explicit that the
   new data is additional and enforce an ordering which is needed for
   predictable `augur merge` behaviour. The canonical data can be
   removed / replaced via step (1) if needed.

I considered adding additional data after the subtype-filtering step,
which would avoid the need to add subtype in the metadata but requires
encoding this in the config overlay. I felt the chosen way was simpler
and more powerful.

Note that this workflow uses an old version of the CI workflow,
<https://github.com/nextstrain/.github/blob/v0/.github/workflows/pathogen-repo-ci.yaml#L233-L240>
which copies `example_data`. We could upgrade to the latest version
and use a config overlay to swap out the canonical inputs with the
example data.