Added monte carlo engine

[nrupprecht]

Added some support for monte carlo.

• Added StochasicProcess as the base class for (loosely speaking) things that can be evolved by the monte carlo engine.
• Added a StochasticProcesse that wraps around existing Diffusion1Ds, added BlackScholes and Heston specific stochastic processes.
• Created MonteCarloEngine, which generates the random variables, evolves the states of the stochastic process it manages, manages the trajectories, etc.
• Added unit tests showing how to price some simple options using different types of stochastic processes.
• Created the AdditionalState class to allow additional data to be tracked during the monte carlo, e.g. the maximum so far, whether a barrier has been crossed by some state variable, realized var, maximum return, etc.

Current limitations:

• The monte carlo engine takes a single stochastic process. This process can have a state of arbitrary size, and can need an arbitrary number of random variable inputs, but all correlations would currently be internal to the stochastic process. This may not be a real flaw, perhaps the right thing to do is always to just have the MC generate uncorrelated random variables of different types, and the model layer determine what the separate pieces of the model are, and how to correlate them.
• It would be ideal to have some abstraction of "instruments that are priceable via monte carlo" and to incorporate the behavior of StochasticProcess more deeply into the existing framework.

[jkirkby3]

We should have a parameter mu which describes the drift rate, and mu = r - q (r = interest rate, q = div yield)

[jkirkby3]

np.ndarray instead of np.array?