add LBG templates into the template set

[moustakas]

z>2 Lyman-break galaxies are likely to be QSO contaminants so we should develop and include appropriate templates for this class of objects. My instinct is that the ELG templates may be sufficient but we should check. Below is a thread between me and Khee-Gan with some relevant ideas.

[John]
This is somewhat out of date, but here's a basic description of how I
construct the ELG templates for the DESI spectral simulations --
   https://desi.lbl.gov/DocDB/cgi-bin/private/RetrieveFile?docid=870;filename=elg-templates.pdf;version=1

Since I'm fitting population synthesis models using photometry that
extends into the rest-frame UV, we should be able to check whether the
sample of DEEP star-forming galaxies are LBG-like in color-color
space.  If not, we'll need to think about how to best include a
dedicated set of LBG templates (do these exist?).

[KG]
  I see that your current templates only extend down to 1250A, which isn't quite 
blue enough for z>2 LBGs with the DESI coverage. In principle we need to go 
down to the Lyman limit at 911A in order to cover the full spectral range that 
might be observed.

  I believe that LBG templates are usually based on the usual families of spectral 
libraries (Bruzual+Charlot etc) but need to be modified to include IGM and dust 
attenuation. The main reference for bright LBGs that I'm aware of is the work by 
Fuyan Bian on Bootes, and BC + IGM + dust is what they assumed:
http://adsabs.harvard.edu/abs/2013ApJ...774...28B

[John]
Note that the QSO target selection also includes a morphology cut --
are the bright LBGs going to be compact enough to pass this cut?  This
is another thing you could help us check...

[KG]
Bian et al 2012 (http://adsabs.harvard.edu/abs/2012ApJ...757..139B) discuss a R=22.3 
LBG that has r_eff=0.3", and in the 2013 survey paper they assumed that their LBGs 
would be unresolved by their ~1" LBT imaging. So I would say that at least some 
(if not most) bright LBGs would sneak past the morphology cut.

FYI I've already pinged Bian for more info but haven't had a response yet. I would guess 
that the most useful thing at this point from him would be the spectral templates he 
used for his selection so that we could translate that into our QSO selection.

[moustakas]

Turns out that the ELG templates go down to ~500 A, so they should be usable out of the box to simulate LBG spectra. However, we need an IGM attenuation module in desisim (or we need to co-opt something from https://github.com/igmhub).

[michaelJwilson]

Not to add further noise, but I have the 5 Shapley composites with lambda = [899., 2000.] AA, split by equivalent width. These could be similar in spirit to e.g. the galaxy PCA, assuming a Fv = constant redwards extension.

I've a script that turns these into gal_maker style (wave, flux, meta), randomly sampling from hardcoded redshifts and colors. I can also turn them into old style rr templates. Is it enough to drop in a list? At what stage in processing does gal_maker start? I also have the same for BC03.

Thanks!

Mike