Allow configuration of rebalance retry strategy on HighLevelConsumer

[cesaraugustogarcia]

• This proposal is to address HighLevelConsumer leaving partitions without owner on multiple consumers startup #541 (HighLevelConsumer leaving partitions without owner on multiple consumers startup). After doing several tests and observations, I believe this (bad) resulting scenario may be caused by race conditions of several consumers trying to consume same partitions and library not handling correctly the retries. Therefore the idea is to allow the rebalance retries to be configurable for each scenario. Particularly, I got good results when setting each consumer with a different (randomised) cycle of retry, to avoid collisions on attempts.
• The change is not breaking at all.
• Also is included a bump on retry library (used on rebalancing action) to version 0.7.0 which introduces a couple of bugfixes. The most important fix introduced is Should set _operationTimeout timeout *before* invoking _fn() tim-kos/node-retry#12

[hyperlink]

Thanks for the PR.

[hyperlink]

Can we change this to the latest and use a ^ instead?

[cesaraugustogarcia]

Absolutely. I just bumped to 0.7.0 trying to be conservative. But I have just checked the latest version, 0.10.0 and has no breaking changes.
So updating my fork changing the commit that has that bump on the retry dependency.

[cesaraugustogarcia]

Alright @hyperlink version changed.