improve the behavior of needs_merge

[evanmcc]

A few things here:

• We shouldn't even call needs merge outside of merge windows, as it's pointless, repeated work and blocks the vnode, raising tail latencies.
• We should add some randomness to the scheduling of needs merge to make lockstep less likely.
• Reconsider the 3 minute needs_merge interval. We should at least make it tunable.

[evanmcc]

From #42, we should also make sure that it's possible to stop the merge worker from doing anything before kv starts up, if possible in a generic way.

[evanmcc]

Going to go ahead and close this, since I think that @engelsanchez's latest changes cover this one. @engelsanchez please reopen and move to 2.1 if there's still work to be done.

[engelsanchez]

I had left it open because my change doesn't address not calling needs merge outside of the merge window. But given that it sounds like a marginal optimization anyway, I'm fine with leaving this closed and re-open in the future if latency tuning says it should be changed.