changes p.stdout.read to communicate()

[samgdotson]

This PR closes #49 by updating fileutils.run() with subprocess.Popen.communicate() rather than Popen.stdout.read() (and similar).

Currently, watts runs external programs by creating a child process with the subprocess package and read/writes the output to the terminal (and a temporary log file) using subprocess.Popen.stdout.read() and .write(). According to the subprocess docs, this can result in a deadlock. The safer approach is to use communicate() instead.

As explained in #49 this issue was discovered because watts deadlocked anytime it ran the abce code. The fix in this PR seems to preserve the intended usage of watts after some ad hoc test runs with abce. However, there are still some issues with abce itself, and since abce is the only code I have access to (within the watts ecosystem), I haven't been able to do more exhaustive tests. Further, I'm still unsure of the following items:

• does the communicate block need to be wrapped in while loop?
• what should the default wait time be? If any at all?