Reduce flakyness of testExemplarsClassicHistogram

[jaydeluca]

I've been encountering issues with io.prometheus.metrics.core.metrics.HistogramTest.testExemplarsClassicHistogram when running tests locally, lot's of these:

[ERROR] Tests run: 28, Failures: 0, Errors: 1, Skipped: 0, Time elapsed: 1.279 s <<< FAILURE! -- in io.prometheus.metrics.core.metrics.HistogramTest
[ERROR] io.prometheus.metrics.core.metrics.HistogramTest.testExemplarsClassicHistogram -- Time elapsed: 0.055 s <<< ERROR!
java.lang.NullPointerException: Cannot invoke "io.prometheus.metrics.model.snapshots.Exemplar.getValue()" because "actual" is null
        at io.prometheus.metrics.core.metrics.TestUtil.assertExemplarEquals(TestUtil.java:13)
        at io.prometheus.metrics.core.metrics.HistogramTest.testExemplarsClassicHistogram(HistogramTest.java:1054)

attempting to make it a little more reliable with awaitility.

Alternatively we could refactor the ExemplarSampler to take a clock as an argument so we could mock it, but I figured we could start here before we change the production code