REF: unnamed pipe -> named pipe (FIFO) for SubprocessTao

[ken-lauer]

Context

Based on #95 as that PR cleans up a lot of subprocess-related handling. It needs to be decided on/merged first.

• We found instability when using SubprocessTao in multiprocessing.Pools
  • Occasionally this would raise OSError: Bad file descriptor (2-10% of the time)
  • It seems as if there are race conditions with os.pipe() and inherited file descriptors in subprocesses, such that they can fight with one another. It never happens on the first initialization from what I've seen.
• Utilizing named pipes, as in this PR, appears to not have these issues
  • In multiple local runs, this has had a 0% failure rate (1000s of runs, 10 multiprocessing subprocesses)
• I think this is a bit cleaner of a solution
  • Even if we figured out how to address the pipe() issue, I think I'd like to move forward with this anyway
• Bonus: custom commands for SubprocessTao
  • Run importable callables in the subprocess and get the result by way of tao.subprocess_call(function, **kwargs)
  • The tao keyword argument is inserted automatically

[ken-lauer]

This is working well in my own tests and also a suite of much heavier testing by @electronsandstuff.
It's ready for review.

If all looks good and we merge this, I think it's about time for a new tag!

[electronsandstuff]

OK, I have run a multiple core example with the modified SubprocessTao class doing the heavy work for eight hours now with good results. The new version seems to be working well.