[SPIKE] Refactoring

[domoscargin]

Some hasty, untested refactoring:

• Moved error handling into its own helper file. The cut is not clean - there's some bits of error handling in the script itself which I didn't get round to factoring out
• Moved octokit into its own helper and created some functions for the actions we want to do. Again, it's not the cleanest cut
• Batch write data to disk to avoid lots of I/O operations
• Create a function for analysing a single repo, then call that function in the loop

I'm pushing and testing this now and I'm pretty sure it'll fail because there's quite a lot of changes. Let's see how it goes!

fixes #12

[domoscargin]

Things I want to do:

• Investigate using a GraphQL query to grab the repo metadata and relevant package.json/yarn/lockfiles in one fell swoop.
• Better separate Octokit from the main script
• Handle errors better
• Figure out how to handle multiple package.json files (in a monorepo, for instance)
• Streamline lockfile reading (we only read it when we know there's going to be a packages entry)
• Take the logic I have for getting parent dependents and add 'em to the output data
• Ditch streaming