Refactor brigade-project-index loader to use a gitsheets API

[themightychris]

This script makes really limited use of gitsheets as a node module to materialize some tables from evolving online data: https://github.com/codeforamerica/brigade-project-index/tree/loader

This illustrates a major early use case for gitsheets though: as a data backend for civic tools and services.

This is proof-of-concept level code, and it does a lot of stuff directly with git that could be made a lot more pleasant with some of the plumbing wrapped up in an expressive gitsheets module API

• Get set up to run and debug the loader locally
• Use gitsheets api to generate more useful commit messages on the index like 🔁 added 1, updated 3 projects from Open Savannah (this example might be accomplished by passing a template for the auto-commit-message generator or making a call to get the generated to use in a sep commit command)
• Make other revisions to taste to the public gitsheets module API to make for more readable code in the consumer (the loader script)
• Ship a minor version bump to the gitsheets backend to npm adding module API enhancements
• Open a PR against the loader inside the cfa repo

[timwis]

So unless I'm underestimating the goal of the brigade-project-index project, I think we only need a couple changes to make it possible with the current high-level API:

1. Support importing JSON data (right now only CSV is supported; if you convert organizations.json to CSV it works fine)
2. Allow multiple datasets per repo. In this case, there will be an organizations dataset plus one dataset per organisation, in projects/{organization}/.
3. Generate more descriptive commit messages, but this is a nice-to-have anyway, I imagine.

At the moment the consumer API looks like this:

const gitSheets = await GitSheets.create()
await gitSheets.setConfigItem('master', 'path', '{{name}}')

const orgsStream = fs.createReadStream('organizations.csv')
await gitSheets.import({
  data: orgsStream,
  dataType: 'csv',
  parentRef: 'master',
  saveToBranch: 'master'
})

It's probably worth sorting out #32 while we're at it but this is the current state.

I think you needed to do all the low-level stuff because (a) json import isn't supported, and (b) multiple datasets per repo isn't supported.