Develop best practice for streaming RPC's
See comment for background info on gRPC buffer: #1502 (comment 160410465)
The gRPC framework used in Gitaly batches data into an internal buffer before sending/flushing to the network. There are a number of RPC's in Gitaly that "chunk" or paginate batches of responses together before sending on a gRPC stream. Since gRPC already does this for us, the chunking in the Gitaly code is redundant and adds unnecessary complexity. Removing these unneeded chunking patterns and advocating a best practice moving forward could help simplify Gitaly RPCs.