Enable users to use different survey setups (pixel_scale, psf_fwhm)

[mr-superonion]

This PR enable users to use different survey setups. related to #188

[mr-superonion]

It is still a draft.

[mr-superonion]

@esheldon, @beckermr , can you please review this PR? It enable users to use different survey's pixel scales and magnitude zero points.

[mr-superonion]

Hmm, I think I need more tests before merging... now the code can make simulations for two observations (HSC: left; DES: right) with the same input galaxies. But it seems that there is a shift between the input galaxies on the sky for these two sims.

[mr-superonion]

Let me add a test running the ML detection code and check the centroids have the same sky coords or not..

[mr-superonion]

I think now the sky coordinate matched between the two sims:

Right is HSC; left is DES.
I would say (hmm...) this is probably ready to merge...

[esheldon]

I like these changes generally.

However, it is really exposing limitations of the design choices we made. For example the Layout objects are nice, and probably we should have used them from the beginning. Then we would not need to pass around coadd_dims everywhere.

It may be we eventually want to do a restructuring around survey objects and layout objects. That would, for example, make it easy to simulate CCDs where coadd_dims is not relevant. Also the survey would carry around pixel scale etc.

[mr-superonion]

make it easy to simulate CCDs where coadd_dims is not relevant.

Thanks!
Hmm, currently, we are using coadd_dims and coadd_pixel_scale, sky_orgin (not implemented, now it is fixed to (WORLD_ORIGIN)) passed to Layout object to define the galaxy distribution (taking flat-sky approximation). At the same time we are using coadd_dims and pixel_scale to define the wcs of the coadd image.

Some possible changes are listed as follows:

• use coadd_length in units of arcmin to define the Layout object and remove coadd_dims from the input.
• change the make_coadd_dm_wcs(layout, coadd_pixel_scale, coadd_dims) to output a set of wcs for coadd pathches.
• enable users to input wcs of single CCD.

[esheldon]

Sorry, I was just discussing these as future possibilities, not for this PR