Error in delta functional approximation of Kelner et al PionDecay implementation

[55hyan]

Hi, I found something wrong in your core, maybe it is not a big problem, but I want to tell you.

(http://naima.readthedocs.io/en/latest/_modules/naima/radiative.html#PionDecay)

In this modules, in def _delta_integrand(self, Epi), you return qpi / np.sqrt(Epi2 + self._m_pi2). But in Kelner et al. 2006, in formula(78), it should be qpi / np.sqrt(Epi2 -self._m_pi2), you should change the "+" to "-".

[zblz]

@55hyan - You are correct, that sign is wrong! Note that this implementation is not the one used in the PionDecay class, which uses the Kafexhiu et al. parametrization.

Fixing the sign resulted in a change of 0.06% over the luminosity of a broad spectrum but might be more important when considering the spectral shape at low energies.

Thanks for the report!