Add support for hedged requests

[magro]

From http://blog.acolyer.org/2015/01/15/the-tail-at-scale/:

Hedged requests: send the same requests to multiple servers, and use whatever response comes back first. To avoid doubling or tripling your computation load though, don’t send the hedging requests straight away:
defer sending a secondary request until the first request has been outstanding for more than the 95th-percentile expected latency for this class of requests. This approach limits the additional load to approximately 5% while substantially shortening the tail latency.

[magro]

Note that there may be different kinds of queries that have different latency histograms (e.g. a plain search, a faceted search or a spelling suggestions). They need to be differentiated when deciding to start a second request.