Allow nimble create to work using CSV input?

[bbimber]

The current workflow is:

• nimble generate
• do manual editing of a JSON file
• nimble compile
• nimble align

there is no good reason step 2 should exist. I'd propose:

• nimble generate <FASTA | CSV | FASTA+CSV>
• nimble compile
• nimble align

In the first step, the input patterns could be:

• Just give a FASTA. This implies you have essentially no metadata beyond the name, which is a valid use case.
• Alternately, you give it a CSV alone. This CSV has a required 'name' column and a required 'sequence' column. In a simple case, the sequence is a literal string. Any other fields in the CSV become metadata
• A version of the above is to give it a CSV with a sequence field. This can either be a literal string, or something formatted like a URI. if that matches genbank://XXXX, we download (lazily during the compile step) the actual sequence. Note: this URI format should allow a substring, like genbank://XXXX:23-2039, which tells nimble to subset the input.
• In a final case, the user provides a CSV and FASTA. Maybe we allow the user to omit 'sequence', and in this case we require the FASTA to have a record where the header matches the 'name' of the CSV?

[bbimber]

One aside: my thinking is that the CSV format would make a friendly human-editable way to prepare and augment the library+metadata. Assuming we can reference genbank in this CSV, we would not need to store the literal sequence, so file size should be relatively small, and suitable to check into repos. The user (or an automated test) would take one of those inputs, run nimble to hydrate it and make the compiled JSON, and then align.