Writer degrades to busy polling after reconnection

[anarthal]

After a successful reconnection, the writer task revers to doing busy polling, rather than waiting for notifications, yielding in a high CPU usage. This is due to the following lines changing the expiry of the writer timer:

redis/include/boost/redis/connection.hpp

Lines 445 to 447 in 328ad97

	// It is safe to use the writer timer here because we are not
	// connected.
	conn_->writer_timer_.expires_after(conn_->cfg_.reconnect_wait_interval);

Timers are stateful, and keep the set expiration time. When a wait is issued against a timer with an expiry date in the past, it completes immediately, hence yielding the aforementioned busy wait situation.

I discovered this while implementing the tests for #250. The fix is trivial - using a separate timer for waits. This also simplifies the logic of #250. I will issue a PR shortly. I've opened the issue to keep track of it, in case anyone has experienced it.