Implementation of a monotone lipschitz network for the MVA functor

The merge request aims to implement a monotone lipschitz neural network for the MVA functor. This kind of network will be used for the Inclusive detached dilepton HLT2 lines and presumably the Topo HLT2 lines. The core of this code is based on the implementation in Allen. It was checked that response of the functor is identical to the original pytorch implementation.

A working implementation of the Inclusive detached dilepton trigger line using this MVA can be found here. Preliminary weight files can be found here.