Flaky testIndexReplay in ThriftElasticsearchTest for es1

testIndexReplay in ThriftElasticsearchTest has been failing occassionally (or most of the time) in the CI build on master.

`Tests run: 35, Failures: 1, Errors: 0, Skipped: 0, Time elapsed: 264.472 sec <<< FAILURE! - in org.janusgraph.diskstorage.es.ThriftElasticsearchTest
testIndexReplay(org.janusgraph.diskstorage.es.ThriftElasticsearchTest) Time elapsed: 16.761 sec <<< FAILURE!
java.lang.AssertionError: expected:<4> but was:<5>`

`The command "travis_wait 50 mvn clean verify --projects janusgraph-es -Pcoverage -Pelasticsearch1,es-docker" failed. Retrying, 2 of 3.`

It happened on [Build #1743](https://travis-ci.org/JanusGraph/janusgraph/builds/461533064), [Build #1709](https://travis-ci.org/JanusGraph/janusgraph/builds/436414375), [Build #1710](https://travis-ci.org/JanusGraph/janusgraph/builds/436416034), and others.

Since this test sub job is in the 'Allowed Failure' category, it may not show up on overall status. Raising this issue to make it be aware of.

Start with testIndexReplay in ThriftElasticsearchTest and reproduce it using the reported Maven command with the elasticsearch1 and es-docker profiles. Compare the intermittent expected 4 versus actual 5 assertion and the listed Travis builds; done means the test passes reliably in the affected CI job.

Written by the indexing model from the issue text.