Add a more general diffusion process

[jorgeecardona]

Would it be possible to add a class with a simple general difussion? Something like $dX_t = b(t, X_t) dt + \sigma(t, X_t) dW_t$ where the functions $\sigma$ and $b$ passed as arguments to the class initialization are both functions of time and space.

I imagine in some circunstances (non-Lipschitz) the behaviour of the solution can't be ensured or easily controlled. But a simple warning in the docs should suffice and maybe some examples of how it fails would be academically interesting for the users.

Best.