test-14-rbart "is reproducible" is not reproducible

[jeffreyoldham]

The recently added "is reproducible" test in test-14-rbart.R can fail depending on the order of random bits. An easy fix would be to set the seed at the beginning of the test, before rbart_vi calls. A better fix would be to test so any random bit order would permit the test to pass.

More details: Tests may be executed in various orders by test harnesses so assuming tests run in a particular order is likely to lead to failures. Running the test by hand in isolation does work for me. Running inside a test harness does not.

Some of the tests in this file do correctly set a seed at the start of the test. Others do not.