Fix Flappy Stats Test

[thanodnl]

This PR is in the hope to fix a flappy test that happens more often than not:

--- FAIL: TestProtocolStats (5.09s)
    Error Trace:    stats_test.go:89
    Error:      Should not be zero, but was 0

    Error Trace:    stats_test.go:90
    Error:      Should not be zero, but was 0

--- FAIL: TestStatsTestSuite (5.21s)
FAIL
FAIL    github.com/uber/ringpop-go/swim 9.984s

The reasoning behind the change is that metrics.Meter mentioned in the comments above is snapshotted every 5 seconds and made available to read from. By exactly waiting 5 seconds we might still read the values just before the snapshot is actually populated with information. By waiting an additional second the hope is that we read less 0's.

[coveralls]

Coverage remained the same at 94.668% when pulling a492b9a on test/flappy-stats-test into 19ff2e6 on dev.

[thanodnl]

Ran the unit tests on travis 6 times without any flap on the changed test.

[mennopruijssers]

LGTM. it's sad though that we can't inject a different time in the meter library.

[mennopruijssers]

[motiejus]

The LGTM is in alt-text.

[dansimau]

I think we can optimise this

time.Sleep(5100*time.Millisecond)