Add consolidated LB usage statistics

Some kinds of aggregated information like uptime or listeners response time would be more properly consolidated if they are computed within the LB module. This BP aims to introduce a way to expose this metrics in order to avoid ceilometer to get huge amounts of samples to achieve fast response to pools or listeners state changes. The initial proposal is to combine up/down events, notifications and smart data grouping to provide:
* Usage time (uptime) in hours
* Aggregated response times in milli seconds