migrate to histogram

[myaser]

Motivation for changing prometheus.SummaryVec to prometheus.HistogramVec:

1. prometheus.HistogramVec can be aggregated, which allows knowing the latency calling a specific node as observed by every other node in the cluster histogram_quantile(0.9, sum by (type, le) (rate(kubenurse_request_duration_bucket{type=~"path_ip-.*"}[5m]))) while this is not possible with prometheus.SummaryVec
2. histogram calculates quantile on the server side, which has 2 positive effects in large clusters.
   1. you can form 0.9 or higher quantile with a lower number of observations per pod KUBENURSE_CHECK_INTERVAL, since the observations made by all nodes can be aggregated together
   2. cheaper observations made by the clients as they do not need to calculate quantiles

for more details

[coveralls]

Pull Request Test Coverage Report for Build 3410042374

• 9 of 9 (100.0%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage decreased (-0.04%) to 77.679%

---

Totals
Change from base Build 3368223944:	-0.04%
Covered Lines:	435
Relevant Lines:	560

---

💛 - Coveralls

[djboris9]

lgtm. @zbindenren how about you?

[zbindenren]

Thanks for your contribution.