Slow fitting (also, clean up timing information)

[bcdaniels]

Fitting of the example data in simpleExample.py is now slower, taking ~10 hours on my laptop instead of "a few" hours claimed in the original. Comparing minimization times stored in minimizationTimeSecondsList, they are indeed roughly 5x longer. We may need to check that integration is actually being done using compiled C code or whether there is a fallback to python code.

Diagnostic timing information is saved when running fitting (in minimizationTimeSecondsList and ensTimeSecondsList) but we may want to revisit this and make it more interpretable—e.g. the typical time it takes to integrate the equations once. Also, the ensemble time appears to be incorrect when running in parallel.