Add initial_image input type

[rmjarvis]

This PR adds the ability for the config processing to read in an initial image from a file and draw additional things onto that. The implementation is pretty simple:

input:
    initial_image:
        file_name: inital_image_file.fits

My test for this is to simulate a supernova time series. It draws a single reference image, and then draws a point source with variable flux 12 times. The SN appears between the 3rd and 4th image. And then it decays exponentially.

[jmeyers314]

LGTM. Just a have few points around my own curiosity; nothing for you to address in code.

[jmeyers314]

Is this needed for the np.exp(-time/50) bit down below? I know import numpy as np is canonical, but I'm a little surprised np.exp works instead of numpy.exp given this line. Just curious how this works.

[rmjarvis]

No. I'm not sure why this is here. I think it must be a relic from something that since changed. I removed it.

But to answer your question about np.exp, we automatically import numpy, both as numpy and as np, so those are always available without anything special on the part of the user.

[jmeyers314]

Also just curious: do yaml documents #2 and #3 automatically inherit the contents of document #1?

[rmjarvis]

Yes. This is kind of a different way of doing templates where everything is self-contained in one file. The first document is the base for everything else. Then each other document starts with that as a template and adds modifications. It's described in some detail in demo6.yaml. Then also used by demos 8, 9, 12.