Uncertainties Scaling Improvements & Info gain debugged (fixed)

[AdityaSavara]

This branch has a fairly substantial set of improvements. Several arguments have been added to UserInput, a few if statements to the code, and some bugs fixed.

1. Uncertainties Scaling had a bug in the scaling of covmat for the prior. Now fixed.
2. Uncertainties scaling can now be turned off, by parameter_estimation_settings['scaling_uncertainties_type'] = 'off'
3. Uncertainties scaling can now be done by a fixed constant by using a string that is a number, like this: parameter_estimation_settings['scaling_uncertainties_type'] = '1e3'
4. Info gain had a bug due to posterior scaling to evidence having an error that was introduced during conversion to log_posterior. That bug has been fixed.
5. A new variable has been introduced called parameter_estimation_settings['mcmc_info_gain_cutoff'] . If that variable is used, by setting it to 1E-4 for example, then small posterior and small prior values will be excluded in the infogain calculation when their pdf is below that threshold.
6. A minor feature was added to complement feature 3. parameter_estimation_settings['undo_scaling_uncertainties_type'] = True will undo a fixed scaling number 3 (above) before returning the prior in getPrior. This currently only works for the case of a fixed constant, so is not normally useful. Maybe it's possible to extend it to the "std" scaling, but it's not that obvious how to do so and I won't worry about it.
7. I merged the Langmuir example into this branch before making the pull request to master, so that example is added to master in this merge also.