Multiplex Bigtable connections over a single grpc Channel

[data-sync-user]

Like a task included in https://github.com/mozilla-services/syncstorage-rs/issues/184,

We currently establish individual grpc channels per client connections for simplicity sake, however we should follow other client implementations and utilize grpc’s multiplex feature where it can support multiple, concurrent requests over a single grpc connection: https://cloud.google.com/bigtable/docs/connection-pools

This will result in:

• More efficient usage of resources, only requiring the overhead of one grpc connection per 100 client requests
• A more well behaved/in line with other client impls

┆Issue is synchronized with this Jira Task