thompson: Generic Multi-Armed-Bandit Simulation
In dfcorbin/npbanditC: (Contextual) Multi-Armed Bandit Simulations

Thompson sampling algorithm is used in order to make decisions. With this function you can choose the mean-reward functions, as well as the basis used to model them.

1	thompson(nsteps, means, sd, basis, posteriors, bas_type)

`nsteps`	The number of steps in simulation.
`sd`	A vector which contains the standard deviation of the reward from each lever.
`posteriors`	A list of posterior distributions. Each posterior distribution is itself a list of model parameters.
`bas_type`	Either "fourier" or "poly".
`mean`	A list, each element should be a function representing the mean for one of the levers available to be interacted with.