Improve ability to use private data

[jameshadfield]

The workflow improvements in #64 enabled both merging of inputs and running in external analysis directories. We've added a placeholder section in the README however there are a number of improvements we could make:

• How can we avoid using accession as the unique ID?
• As private data is uploaded to PPX could we identify duplicate sequences?
• What columns are needed for different builds - can we check this using code, or is it a docs issue?

[victorlin]

How can we avoid using accession as the unique ID?

Maybe nextstrain/augur#1780? Augur would expect an id column on the merged data. In the workflow, we'd create id as a copy of accession for Pathoplexus data. Private data can provide id in whatever fashion.

[victorlin]

After working on nextstrain/augur#1780 I'm thinking it's better to make the id column configurable per input. Example:

inputs:
  - name: ppx_open
    species: bdbv
    metadata: "s3://nextstrain-data/files/workflows/ebola/bdbv/metadata_open.tsv.zst"
    id_field: accession
    sequences: "s3://nextstrain-data/files/workflows/ebola/bdbv/sequences_open.fasta.zst"

additional_inputs:
  - name: private
    species: bdbv
    metadata: metadata.tsv
    id_field: custom_id
    sequences: sequences.fasta