LQG_path_marginals: Obtain marginal distributions of path measure
In jeremyhengjm/SchrodingerBridges: Numerical implementation of Schrodinger bridges

# Function to compute marginal distributions of path measure given initial a Gaussian initial distribution and Gaussian Markov transition kernels of the form: x_t | x_t-1 is Gaussian with mean = K_t x_t-1 + r_t and covariance = S_t. "initial" is a list containing the initial parameters. "transitions" is a list containing Kmats, rvecs and Smats, which are arrays containing the transition parameters. For each, the range of the last index gives the number of steps.

1	LQG_path_marginals(initial, transitions)

`initial`	list with keys: `mean` `cov`
`transitions`	list with keys: `Kmats` array of size d x d x T `rvecs` array of size d x T `Smats` array of size d x d x T