Optimize BigtableClientImpl pooling

[data-sync-user]

As part of https://github.com/mozilla-services/autopush-rs/commit/2fd708654e92838953d030274bbf3de52b45f7d0 I decided to construct the Pool to contain BigtableDb, which is owned and managed by us. While this works, it may be more efficient to pool the grpcio::Channel tjat BigtableClientuses. Neither Channel nor BigtableClient are owned by us, so they’re subject to change, however there is a BigtableClient::new(channel) function which we might be able to use to create the client. Ensuring that the channel is available to extract, however, might be trickier.

┆Issue is synchronized with this Jira Task