Automatic differentiation support in volesti

The most efficient algorithm to sample from a log-concave distribution, that volesti supports is the Hamiltonian Monte Carlo with leapfrog integration. The leapfrog method is an Euler method that in each step evaluates the gradient of the log-probability density function of the target distribution. To sample from a log-concave distribution volesti currently requires a function that evaluates this gradient as an input. However, there are plenty of applications that require sampling from a distribution where the gradient of the log-probability density function is unknown. This project will address this problem and will provide to volesti routines of automatic differentiation of a multivariate function. Then, a function that evaluates the log-density will suffice to sample from the target distribution.