refactor tests / testing approach

re: https://github.com/shaunlebron/parinfer/issues/168

I would like to move away from the "test cases in markdown files with custom parsing" approach.

I think the custom annotation syntax is valuable for seeing how Parinfer operates. The way it is currently parsed and used by the test suite is pretty opaque right now; I think the implementation could be cleaner.

Implementation TODO

• separate the annotation syntax parsing / reconstruction into it's own library with dedicated test suite (possibly a separate repo?)
• create a kidif file for each test case
• give each test case a stable id (use the 1000's for indent mode, 2000's for paren mode, 3000's for smart mode)
  • done with PR #14
• write a script that builds a markdown file from the test cases for readability / exploring how Parinfer works