feat(spike): rover core build

[EverlastingBugstopper]

This PR introduces a YAML config format for specifying multiple subgraphs to compose.

example YAML with two subgraphs titled films and people:

subgraphs:
  films:
    routing_url: https://films.example.com
    schema: 
      file: ./good-films.graphql
  people:
    routing_url: https://people.example.com
    schema: 
      file: ./good-people.graphql

$ rover core build --help
rover-core-build 0.0.3
Build a core schema from a set of subgraphs

USAGE:
    rover core build [OPTIONS] --config <core-config>

FLAGS:
    -h, --help       Prints help information
    -V, --version    Prints version information

OPTIONS:
        --config <config-path>    The relative path to the core configuration file
    -l, --log <log-level>          [possible values: error, warn, info, debug, trace]

This will attempt to compose your subgraph schemas. If it successfully composes, it will print the composed CSDL. Paths are computed relative to the config file itself. We could eventually look for a rover.toml by default, but for now, since this is the only config file we have and nothing has been set in stone for other commands, we require an argument that takes a relative path.

Example:

$ rover core build --config spike/composes.yaml
CSDL:
schema
 @graph(name: "films", url: "https://films.example.com")
 @graph(name: "people", url: "https://people.example.com")
 @composedGraph(version: 1)
{
 query: Query
}

directive @composedGraph(version: Int!) on SCHEMA

directive @graph(name: String!, url: String!) repeatable on SCHEMA

directive @owner(graph: String!) on OBJECT

directive @key(fields: String!, graph: String!) repeatable on OBJECT

directive @resolve(graph: String!) on FIELD_DEFINITION

directive @provides(fields: String!) on FIELD_DEFINITION

directive @requires(fields: String!) on FIELD_DEFINITION

type Film {
 id: ID!
 title: String
 actors: [Person]
 director: Person
}

type Person
 @owner(graph: "people")
 @key(fields: "{ id }", graph: "people")
 @key(fields: "{ id }", graph: "films")
{
 id: ID!
 name: String
 appearedIn: [Film] @resolve(graph: "films")
 directed: [Film] @resolve(graph: "films")
}

type Query {
 film(id: ID!): Film! @resolve(graph: "films")
 films: [Film] @resolve(graph: "films")
 person(id: ID!): Person @resolve(graph: "people")
 people: [Person] @resolve(graph: "people")
}

and

$ cargo run -- core build --config spike/errors.yaml
error: KEY_FIELDS_MISSING_EXTERNAL: [films] Person -> A @key directive specifies a field which is not found in this service. Add a field to this type with @external.
error: KEY_FIELDS_SELECT_INVALID_TYPE: [films] Person -> A @key selects idd, but Person.idd could not be found
error: KEY_NOT_SPECIFIED: [films] Person -> extends from people but specifies an invalid @key directive. Valid @key directives are specified by the originating type. Available @key directives for this type are:
        @key(fields: "id")
error: Encountered 3 composition errors while composing the graph.

---

Error cases we need to handle better we currently handle:

• Config file cannot be found
• Config file cannot be read to a string
• Config file cannot be parsed
• Schema path defined in the config file cannot be found
• Schema file cannot be read to a string

^ These are all handled now!

• Probably others I'm not thinking of

^ I'd love some QA on this to see if folks can break it and we can add more test cases 😄

---

It's important to note that we do absolutely no input validation on the URLs due to the fact that I think they can really be anything, even empty strings. That being said, we do require that it is defined for each subgraph.

There are three options here:

• Change nothing in this PR, allow any string to be passed, but do not parse it or require it to be well-formed.
• Do not require a routing_url for each subgraph and pass an empty string to harmonizer if it is not specified in the config file.
• Be more strict/opinionated about routing_url and throw a parse error if a routing_url does not exist or if it's not a well-formed URL that would produce a bad artifact.

Closes #350

[EverlastingBugstopper]

from @ndintenfass in Slack:

This looks great - I propose we change schema_path a bit, either to shorten it in place, or maybe something that has more flexibility like:

subgraphs:
  films:
    routing_url: https://films.example.com
    schema:
      file: ./good-films.graphql
  people:
    routing_url: https://people.example.com
    schema: 
      file: ./good-people.graphql 

That would let file become something like graph or introspection_url (or even raw per earlier discussion that if all of them resolved to the raw internally people could build their own tooling in a pinch)

for instance, then I could later do something like:

subgraphs:
  films:
    routing_url: https://films.example.com
    schema:
      file: ./good-films.graphql
  people:
    routing_url: https://people.example.com
    schema: 
      graph: mygraph@current
      subgraph: good-people

[lrlna]

Let's bring this in as part of 0.0.4, so folks can try it out before our release next week.