Use the Fermi-Dirac distribution for efermig instead of Methfessel-Paxton smearing in EPW

WARNING: Please do not accept this merge request before Tues. 28 April so that other developers can have a chance to comment on this.

We made the following substitution for the "efermig" calls in EPW: "ngaussw"->-99 (which corresponds to a Fermi-Dirac smearing) and "degaussw"->"eptemp" (which correspond to evaluate the Fermi-Dirac function at the "eptemp" temperature).

We made the same substitutions the following sections of the code, namely elph2.f90 ephwann_shuffle.f90 ephwann_shuffle_mem.f90 io_eliashberg.f90 io_transport.f90 selfen.f90 and utilities.f90

Francesco Macheda and Samuel Ponce