Increase batch timeout default for client

Changing this default to 0.2s tends to completely mitigate issues seen in !18 (merged). My guess is that selecting a timeout close to the round-trip time is causing the client to needlessly reconnect often. Why this causes a memory increase is beyond me, but regardless, the new default should be a reasonable tradeoff.

Also document the batch_size and batch_timeout options in the client.

Closes #18 (closed).