Grep is working inefficient for test cases with a large number of files

[1999]

@createthis in #203

I do notice that this feature seems fairly inefficient with 1100+ tests spread out amongst 93 files. I get a lot of one line outputs and it takes much longer than a regular mocha run. I assume this is because it is loading mocha in a separate process for each file, then grepping, and finally returning and compiling all of the results into the output. An optimization might be to pre-grep the files and only spawn processes for files that appear to have matches.