Add configurable timeout on gazctl gRPC calls

[rupertchen]

Request

Provide a way to configure the timeout in seconds for gazctl calls.

Background

If the address being dialed by gRPC is invalid, it will retry indefinitely (for a very very long time) in the background effectively hanging the tool. This is frustrating for the operator as they may wait a long time believing the server is just taking longer than usual.

Proposal

Add a global --dial-timeout that takes the time in seconds to wait to open gRPC client connections. Use a reasonably generous default (10s?) and allow 0s to mean no timeout. This should apply to both gRPC connections to the Gazette cluster or Gazette consumer shards.