Make code and tests flexible to the use of a pre-processed Montgomery dataset to speed-up testing where possible (closes #74)