Bugfix sed

[pmelchior]

SED initialization code was not robust for negative/zero flux in some bands. As such, such an SED is not a problem. I therefore changed the behavior from raising an exception to logging a warning.

I also, moved the SED correction (for PSF width variations) from get_pixel_sed to the relevant init_ functions because this is where it's needed. It's confusing otherwise.

[fred3m]

I like the idea of only using positive values of the sed but I have a design question and a question about a negativity check in my comments below.

[pmelchior]

@fred3m , I've addressed all comments, and created a simpler implementation of MultiComponentSource that actually uses a common center for all components. You can merge if you are fine with it.

[fred3m]

Thanks for making the changes. Everything looks good except that I didn't quite like the change in MulitComponentSource where you removed the update function from the individual components and instead put them in the parents update function. While it works fine this way I think that it goes against our design pattern of having child components with update functions and a parent that just updates the children. Feel free to push back if you feel that I'm being too rigid.