(feature): Add encoding and source nodes

[amckinney]

This updates the OpenAPI and IR generators to include encoding and source nodes where relevant. This makes it possible to track the original source of an OpenAPI type, so that we can include additional context in the generated IR.

For Protobuf inputs, we include a lightweight ProtobufParser to read specific values we're interested in (including the gRPC service name).

In a follow-up, we'll add even more information in the IR's top-level SourceConfig property, and add more configurability via the x-fern-encoding extension.

Note that this also enables LFS for the fhir.json testdata files since they're way too large (hence the significant number of removed lines in this commit).