redis retry can cause exceeding 1 second timeout.

[drewrobb]

With the changes introduced in b96fbb6 we are seeing in production it is still possible for a pump send to timeout. The retry after a first error opens a connection again and tries to push again. If the first error is because of a timeout, there might not be enough time remaining out of the 1 second budget given by logspout. Fix ideas I have are introducing a timeout external from the redis client, simply removing the retry code, or fixing the 1 second timeout budget + closing the pump permanently after a single timeout upstream (I've opened an issue for this).

[rtoma]

Thank you for your comment and the upstream issue.