When the runner contract is specified as "run all the tests in the file in sequence,
resetting cluster after each file", it makes is hard to translate the concept into
regular Test::Unit environment.
Changed the "Analyze API JSON format - index and field" test to perform the necessary
setup as part of the test code.
Possible to fail this test with a 500 NoShardsAvailable error, due
to the index not being fully allocated before the analyze call is
executed.
Changing for a wait_for_status: yellow fixes the occasional problem.