Fix hiredis blocking indefinitely despite timeout being set (#593)

[catterer]

Currently, redisSetTimeout function does not update command_timeout field in the context. The following call to redisReconnect then resets the socket timeout to infinity. This is probably the root cause of the issues (#593), (#718), (#263), (#517).

This PR fixes the bug and also fixes the corresponding test, which had two issues:

1. DEBUG SLEEP 3 was sent twice because manual c->funcs->write did not flush the connection buffer
2. DEBUG SLEEP 3 in one test affected the following test

[michael-grunder]

Thanks for the PR.

I'm in favor of merging this, but I'm wondering whether this is technically a breaking change?

Relevant XKCD

[michael-grunder]

Closing in favor of #1093 (which is just this PR rebased against master).