Triple-Gauss PSF format

[cdeil]

I've added a formula for the triple-Gauss PSF: http://gamma-astro-data-formats.readthedocs.org/en/latest/irfs/psf/psf_3gauss/index.html as well as some background info on radial PSFs: http://gamma-astro-data-formats.readthedocs.org/en/latest/irfs/psf/index.html#point-spread-function

I think some discussion / checks are needed:

• [ ] Is the formula correct? I took it from the HAP code that fits the PSF and added the 1/pi factor to go from dP/dr^2 (which is what HAP fits) to dP/d\Omega (which is what Fermi and OGIP use as PSF value).
• [ ] Is there a more sane / stable parametrisation for multi-Gauss (look what PSFEx do) that we want to adopt?
• [ ] Can the presentation / explanations be improved?
• [x] @mimayer - Does the PA exporter use the same parametrisation? ==> there is no 3-gauss exporter in PA.
• [x] @JouvinLea - Do you use multi-Gauss in HAP-FR exporters? Same parametrisation? ===> This is work in progress, HAP-FR can put whatever we decide here.
• [ ] I'd like to check the Gammapy code
• [ ] @mimayer - Is gammalib.GCTAPsf2D consistent with this definition?