[New Feature]: Use object noise and not just sky noise in calculating score image

[wmwv]

Is your feature request related to a problem? Please describe.

Current score image from SFFT is based on the sky noise. We want to include the full object noise to characterize significant. This might actually be an Issue against sfft-romansnpit, but I might play around with it here in these repo before deciding on what changes to implement in a cuda-fied way in SFFT. First I need to spend some time to really think about how well the current score image is doing.

Describe the solution you'd like

1. Use the full variance, including for the shot noise of the objects.
2. Consider using a mildly-blurring kernel to account for WCS/position imprecisions (this probably really matters for ground-based more than space).

Describe alternatives you've considered

Just mask bright stars (#19). Will work.

Additional context

Doing this right would be worth writing up in a little paper. It's a thing we've known we've been doing wrong for a long time, we just haven't cared when searching for faint transients.