feat!: Use workflow output commands instead of check runs

This solves the annoying check-run-mislabled issue, but is a breaking
change: When an action run has annotations to create, by default, it
will now mark the job as successful instead of neutral.

More explanation, discussion, context can be found on this commit
message. We hope to "unbreak" this once Github (re)adds the behavior we
need.

planningcenter/balto-rubocop@7b5adc8

README.md

@@ -52,7 +52,7 @@ jobs:
 
 | Name | Description | Required | Default |
 |:-:|:-:|:-:|:-:|
-| `conclusionLevel` | Which check run conclusion type to use when annotations are created (`"neutral"` or `"failure"` are most common). See [GitHub Checks documentation](https://developer.github.com/v3/checks/runs/#parameters) for all available options.  | no | `"neutral"` |
+| `conclusionLevel` | Which workflow status should be used when annotations are created. Currently, `"failure"` and `"action_required"` show as failures, while everything else (including `"neutral"`) show as successful | no | `"neutral"` |
 | `failureLevel` | The lowest annotation level to fail on | no | `"error"` |
 | `extensions` | A comma separated list of extensions to run ESLint on | no | `"js"` |
 | `dependencyInstallMode` | `"smart"` or `"none"`. Control how dependencies are installed (if at all). Smart (requires yarn) will attempt to install the least amount of packages to successfully run eslint.| no | `"smart"` |