More succinct paralleltest test output.

[kwmsmith]

I've run into problems a few times where the first 1 or 4 parallel engines have errors, but the latter ones don't. The way the output is ordered, the latter tests make it seem like everything is fine.

This makes the output more compact to give a quick overview of the success or failure of the parallel tests; it is always possible to look at the file in .parallel_out for more details.

For example:

$ make test_engines
Running 'python -m unittest discover -s distarray/local/tests -p 'paralleltest*.py'' on each engine...
====>  .parallel_out/2.7.3-unittest.out.1.00
OK
====>  .parallel_out/2.7.3-unittest.out.1.01
OK
====>  .parallel_out/2.7.3-unittest.out.1.02
OK
====>  .parallel_out/2.7.3-unittest.out.1.03
OK
====>  .parallel_out/2.7.3-unittest.out.1.04
OK
====>  .parallel_out/2.7.3-unittest.out.1.05
OK
====>  .parallel_out/2.7.3-unittest.out.1.06
OK
====>  .parallel_out/2.7.3-unittest.out.1.07
OK
====>  .parallel_out/2.7.3-unittest.out.1.08
OK
====>  .parallel_out/2.7.3-unittest.out.1.09
OK
====>  .parallel_out/2.7.3-unittest.out.1.10
OK
====>  .parallel_out/2.7.3-unittest.out.1.11
OK

[coveralls]

Coverage remained the same when pulling 1d5a31b on feature/compact-paralleltest-reporting into 7a7bf40 on master.

[bgrant]

Looks good to me.