Collect better error messages from failures

[rcurtin]

Sometimes failures can give us results like:

Traceback (most recent call last):
  File "/home/jenkins/workspace/pull-requests-benchmarks-benchmark/tests/benchmark_decision_tree.py", line 261, in test_RunMetrics
    self.assertTrue(result["Runtime"] > 0)
TypeError: 'int' object is not subscriptable

with stdout

[FATAL] Could not execute command:
orkspace/pull-requests-benchmarks-benchmark/libraries/bin/mlpack_decision_tree',
'-t', 'datasets/iris_train.csv', '-T', 'datasets/iris_test.csv', '-v', '-p',
'mlpack_dct_predict.csv'] The error is: Command
orkspace/pull-requests-benchmarks-benchmark/libraries/bin/mlpack_decision_tree',
'-t', 'datasets/iris_train.csv', '-T', 'datasets/iris_test.csv', '-v', '-p',
'mlpack_dct_predict.csv']' returned non-zero exit status -6

It would probably be helpful if we could provide the actual stdout or stderr output from the subprocess calls.