I agree with Andrew, that the command line would be a good way to 
interface with the model, especially if configuration files could be 
passed from the command line.

In this commit, I add that feature using a console_scripts entrypoint in 
setup.py to a special new function called run_model, and this function 
runs the model for a given number of timesteps. This number of timesteps 
is set by the config file, which now accepts an additional timesteps 
parameter.

Basically, since we don't get a model instance to work with when we test 
from command line, I just hacked this together to save an output image, 
and check if that image exists. This is enabled with a new 
test_output.yaml file.

I also added a test to make sure the versions match when we query from 
command line.

now execution from a config file can be initiated either from a bash 
entrypoint or from a main file when called via 'python -m pyDeltaRCM' 
This change addresses DeltaRCM#22.

Issue was with the subprocess package. I was using kw arguments that 
were introduced in 3.7, so I reverted the call to the 3.6 version, now 
it should run on both.

@amoodie pointed out some issues with the PR. changed some importing 
code, and one order of operations problem.

Surprise! I didn't know how this worked. Turns out you don't need to 
import the whole package anyway. derp derp