* Restructures the `cabal init` command to fix historical
  issues. All flags are preserved.
  * Codebases for interactive and non-interactive flags
    are disentangled.
  * Data structures now exploit relevant stanza structure
    and formatters only care about stanza data
  * Heuristics and prompts have a pure and impure implementation.

* Sets default behavior to be `--interactive` as opposed to
  `--non-interactive`.

* Rewrites tests to achieve 98% coverage
  * Golden files now test every stanza individually
  * Every flag is covered by a unit test
  * Interactive, simple, and non-interactive workflows are
    covered.