feat(gateway): Implement @core v0.2, including "for" and core__Purpose support

[trevor-scheer]

This PR updates the gateway to @core v0.2 🎉

This includes implementing support for the optional for argument in the @core directive. This argument allows a core directive usage to specify how a gateway should behave in the case that its feature is unsupported. This is important for various use cases - in particular, execution and security concerns.

the gateway will throw UnsupportedFeature on schema update IF:

1. it sees a core feature specified which it doesn't implement AND
2. that core feature provides the for argument (only valid options are currently EXECUTION and SECURITY

additionally, if the supergraph specifies core v0.1, and contains core directives which use for:, the gateway will report ForUnsupported.

TODOs

• MIT license @apollo/core-schema Add MIT license core-schema-js#11
• Release @apollo/core-schema? I don't think we should be depending on a github branch, and certainly not a main branch. apollographql/core-schema-js@e6c318e
• Changelog entry

Closes https://github.com/apollographql/polaris-planning/issues/109

[queerviolet]

this looks good! my only question is about the behavior we want to have if we see for: something in a core v0.1 document. currently, we explicitly do not honor it and fail open, which seems… maybe not great, since we understand the likely intent, and the intent was to protect the schema from being served by us dunderheads who don't understand how the @auth directive works (or whatever)

three options:

1. hard fail if we see a for: in a core v0.1 doc. this will teach people (mostly us, probably ;) ) to update the feature in their docs / generators
2. warn if we see a for: in a core v0.1 doc, but honor its intention (i.e. ensure the feature is supported). this will annoy people into eventually updating the feature, and meanwhile mostly do what they want
3. do not warn if we see for: in a core v0.1 doc and honor its intention

i think (1) is a bit harsh and (3) is a bit loose, so i lean towards (2). wyt?