All 4 call types share these 4 arguments:
Task<Metadata> responseHeadersAsync, Func<Status> getStatusFunc, Func<Metadata> getTrailersFunc, Action disposeAction.
Wouldn't it be useful to wrap them in a single class (I am actually undecided what's better here)?

The main advantage of having a class is future proofing the signature, so if we can add stuff in it in the future (which I think is a worthy goal), but if we are confident about the signature, I don't think wrapping in a class is more user-friendly here. What do you think?

alternatively, we can provide factory methods:
currently, we have 2 factories: https://github.com/grpc/grpc/blob/master/src/csharp/Grpc.Core/Calls.cs (for "actual" calls)
and https://github.com/grpc/grpc/blob/master/src/csharp/Grpc.Core.Testing/TestCalls.cs (for test doubles).
Perhaps we could have a 3rd factory class for "intercepted" calls. That way, we would have an intermediate layer in case we wanted to make changes to Async**Call constructors.

The ChannelExtensions in the interceptor PR acts as the third factory, but I picked the extension route to make the API nicer, i.e. enable channel.Intercept(...).