feat: structured output

[lrlna]

Description

Several of Rover's commands would benefit from structured output. The working assumption is this would be in the form of JSON payloads of a defined shape (and likely with a defined "envelope" where every command offering structured output would have the same top-level keys and a payload specific to the output).

Working assumption is we'd have a flag like --json or --output json

[abernix]

As a use-case from apollo-tooling: apollographql/apollo-tooling#2213:

Sugestion

Would be awesome if we can pass a flag --json to apollo service:list to consume the output or have a programmatic version of apollo service:list to create nodejs scripts based on the output of this command.

Why

I want to create a variant schema based on another variant, to enable things like app reviews. The idea is to enable into PRs to publish an app review of a specific service and compose this service with the current staging schema.

[abernix]

Here's my own framing of a use case that was brought to my attention:

JSON output enables implementors to build their own presentation layer or Dashboard for this data and display it to their developers within existing deployment pipelines. It integrates well with other tools — like Splunk — to generate reports about which subgraphs are failing (and for which reasons) and display overall deployment health at a glance.