gRPC support in .http (GRPC method, reflection, streaming via ===)

[syncended]

Goal

Extend yapi to send gRPC calls using the same `.http` file format, mirroring IntelliJ HTTP Client's gRPC syntax for compatibility.

Status

v0.2 feature. Depends on saved-requests format (#11) being live and parser-tolerant of arbitrary uppercase methods + `# @key value` metadata (already required for v0.1.0).

Syntax

```http

Unary call

GRPC localhost:50051/helloworld.Greeter/SayHello
authorization: Bearer {{token}}

{ "name": "world" }
```

• Method: `GRPC` (use `grpcs://` URL scheme or `# @grpc-tls` for TLS).
• URL: `host:port/<fully.qualified.Service>/`.
• Headers = gRPC metadata.
• Body: JSON, mapped to protobuf via `JsonFormat`.
• Streaming: messages separated by `===` (client/bidi streaming → multiple request messages; server/bidi → multiple response messages).

Schema source

• Default: gRPC server reflection (`grpc.reflection.v1alpha.ServerReflection`).
• Override per-request: `# @grpc-proto ./protos/foo.proto` + `# @grpc-import-path ./protos`.
• Global defaults in `config.toml`:
  ```toml
  [grpc]
  reflection = true
  proto_paths = ["./protos"]
  ```

Tech stack (proposed)

• `io.grpc:grpc-netty-shaded`, `grpc-protobuf`, `grpc-services` (reflection client).
• `com.google.protobuf:protobuf-java-util` for JSON ↔ `DynamicMessage`.

Out of scope

• Code-gen of Kotlin stubs from `.proto` (we use `DynamicMessage` for runtime flexibility).
• gRPC-Web.

Acceptance criteria

• A `.http` file with `GRPC` method executes a real unary call against a server with reflection enabled.
• Local `.proto` path works as a fallback when reflection is disabled.
• Server-streaming, client-streaming, and bidi modes all work; multiple messages are rendered with `===` separator on output.
• Existing v0.1.0 `.http` files for HTTP requests keep working unchanged.