Support Retry Strategies, etc...

The work for this is in LCB, we need to expose it.  I believe we should be able to create a retry strategy in python that can be communicated back to LCB, switch between them, query for what the current one is, etc...

Additionally I'm sure there are some settings we could expose?  Look into that just in case, and read the RFC to be sure I've not missed anything major.  Most of the work is done by LCB, we mainly just need to expose what needs exposing.  

I've put this in 3.0.0, as it is a feature but it is additive at least, so not going to gate RC with it. But it is pretty important to knock out early, so doing it during the RC sprint is ok with me.