Debug the parallelization issues with the ensemble

[jlperla]

See https://github.com/JuliaDiffEq/DiffEqBase.jl/blob/master/src/ensemble/basic_ensemble_solve.jl#L121

and others. Seems like the batch_func is used for both the Serial and Distributed... both of which are typical ways to use julia. Maybe the threads should just be ignored since the implementation is preliminary.

Worst case, use prob_func to do a copy of the data yourself.