Support a multi-process Danger with custom runners

[orta]

OK, now bear with me.

1st, consider this. With Peril I created the ability to see the Danger DSL for any PR:

• https://peril-staging.herokuapp.com/api/v1/pr/dsl?owner=artsy&repo=emission&number=681
• https://peril-staging.herokuapp.com/api/v1/pr/dsl?owner=danger&repo=danger-js&number=338
• https://peril-staging.herokuapp.com/api/v1/pr/dsl?owner=artsy&repo=metaphysics&number=715

I did this so I could make better documentation inside danger.systems, but what would happen if this were formalised a bit, and inside danger instead of Peril?

Imagine that we add a new command:danger process.

e.g. danger process - which would output the above JSON to STDOUT.

Thennnn, we also make a standard in terms of the results. Today, Danger expects results in this form:

{
  "fails": [{ "message": "Failing message Failing message" }],
  "warnings": [{ "message": "Warning message Warning message" }],
  "messages": [{ "message": "message" }],
  "markdowns": ["markdown"],
}

So what about if danger process could also take the above as STDIN. Now you can make a pipeline.

danger process > [ x ] > danger process

It leaves the really gnarly work of handling different CIs, getting all the useful data from GitHub/GitLab/BitBucket/etc, amending/creating/deleting Dangerfiles to DangerJS. A Danger runner just need to take in a DSL as JSON, and output a results JSON to STDOUT.

This means, anyone, in any language could make a Danger runner. So, there can be Swift/Rust/Go/Whatever Dangerfiles in the future.

[orta]

This is in.