Add backoff on connection retries

[vshlapakov]

The PR adds simple backoff logic to add optional delay for connection retries.

Review, please.

[mcclymont]

Hi. Can you please explain why you think this retry delay is needed? Did you encounter a situation where it retried very often?

If needed, I think it would be better to just include a mandatory retry delay of a few seconds rather than adding a configuration option.

[vshlapakov]

Hi @mcclymont Of course, sorry for being too brief.

My usecase is a distributed system with collectd + potsdb when OpenTSDB instance can be down (due to maintenance, for example), and not having backoff on retries looks pretty harmful for the system. I don't mind if this delay will be fixed, as you wish! Usually people want to have an ability to tune everything :)

[mcclymont]

I've implemented a simple exponential backoff instead 0fb0bbf. Released in v1.0.3. Thanks for your input.

[vshlapakov]

Thanks @mcclymont!