New flat histogram parallelization implemented

[HaroldHatchNIST]

In version 0.25.7, flat histogram tutorials used CollectionMatrixSplice with OMP parallelization of macrostate windows. Starting from version 0.25.8, those tutorials now parallelize with Python user control of the macrostate ranges in the Python launch script. To allow all threads to continue running on a node until every thread is finished, files determine simulation termination in the new class "WriteFileAndCheck" and "Run" class argument "until_file_exists." The new parallelization implementaiton splices macrostate probabilities, which leads to the cost that one more macrostate per window must be simulated but benefits from allowing different trials on threads and freeing up OMP threads for other parallelization methods. We still support the CollectionMatrixSplice parallelization method.