Changes

• Update the timeout written in the query request payload when retrying, according to the time remaining until the deadline.
• Fix a bug where not setting the timeout in the options (i.e. relying on the cluster-wide default) resulted in the query timing out immediately.
  • This is because the HTTP component's timeout would not be set, which resulted in the epoch being passed on to the http_session_manager's connect_then_send_pending_op as the deadline.
  • The timeout is now always sent to the HTTP component.
• Remove the dispatch timeout test which was succeeding because of the bug above. Dispatching appears to succeed very quickly, so even with a timeout 1ms the test was not consistently passing - we probably need to find another way to test it.
• Use the default timeouts in tests that don't need a specific timeout value (this was previously hiding the bug)