[FlightRPC] Remove sketchy gRPC reflection code

[lidavidm]

Describe the bug, including details regarding any error messages, version, and platform.

grpc/grpc-java#12579

There are public methods to get ownership of the internal ByteBuffers, instead of having to use reflection (which, did we actually ever need it in the first place?)

There's no need to dig into our internals for accessing the ByteBuffer (which is what GetReadableBuffer reports to do). We have zero-copy APIs. They can use HasByteBuffer.getByteBuffer()+InputStream.skip() to loop through the ByteBuffers. If they want to access all the byte buffers simultaneously (as skip() will deallocate the last byte buffer, just like read() does), then they can use InputStream.mark(). If they want to take over ownership of the ByteBuffers, then they can use Detachable.detach().

[lidavidm]

To anyone who wants to implement this, it would be helpful to show that the end result is zero-copy or demonstrate that it benchmarks the same as before (albeit our hand is forced by grpc/grpc-java#12580)

[ejona86]

If anyone has questions on how to reproduce a particular behavior, I'd be happy to help. I'll probably try to glance at any PR that comes from this, since in the past I've seen people not being too familiar with mark()/skip().

[tadeja]

take

[tadeja]

(Discussed at conference sprints and getting pr ready with @mihazagar)