Adjust Av and Age priors, allow verify_params failures

[lcjohnso]

Here I enable three optional features:

1. Add an exponential Av prior model.
2. Allow the removal of the "constant in linear age" prior to prefer the grid's intrinsic logarithmic spacing.
3. Allow verify_params failures to print to screen, but continue without exiting. This is helpful when using non-fA dust curves and fixed R_V values.

While not included in this PR, I would be happy to add info about what needs to be changed in run_beast.py and datamodel.py to use these updates, but I'll wait until the datamodel.py documentation page is drafted (Issue #71).

[coveralls]

Coverage increased (+0.003%) to 12.167% when pulling 7546400 on lcjohnso:adjust_priors_lcj into 6d93d78 on BEAST-Fitting:master.

[karllark]

In the verify_params changes, I'm not sure having the exit being overridden for all possible exits is a good idea. In fact, ideally the override would not be done at all, but the verify_params logic changed to handle the two cases mentioned - no fA and/or fixed Rv extinction curves. @lcjohnso: Is such a change in the logic not possible/straightforward?

[lcjohnso]

Re: @karllark's suggestion, I added logic to verify_params that allows for non-exit() warnings in the case that:

1. fAs = None, as needed for case where non-Gordon16 extinction curve is used.

2. Parameter grids are defined as single values (e.g., defining sequence is [0.0, 0.0, 0.1] leading to single grid point at 0.0).

I'm leaving the noexit option implemented (could still have use in future), but its use is discouraged as it might allow non-physical grid/parameter values.

[lcjohnso]

Also: I added initial documentation of datamodel.py/run_beast.py, including HowTo instructions for implementing new optional features.

This addition is a good step towards closing Issue #71 -- please comment there on additions that might still be needed before closing issue.

[coveralls]

Coverage decreased (-0.002%) to 12.162% when pulling 0938598 on lcjohnso:adjust_priors_lcj into 6d93d78 on BEAST-Fitting:master.

[coveralls]

Coverage decreased (-0.002%) to 12.162% when pulling 7c17b2d on lcjohnso:adjust_priors_lcj into 6d93d78 on BEAST-Fitting:master.

[karllark]

This should be N*e^-ax.
It would be nice if the 'a' parameter had a variable name that was more descriptive. I can't think of one offhand, but it is not some kind of timescale?

[lcjohnso]

I corrected the notation, but am leaving the param as "a" -- could use "decay_rate" or "decay_param" but that seems overkill. Instead, I added additional text in docstring.

[karllark]

Is this parameter needed? Shouldn't this be set by the desire to have this prior integrate to 1 over the min/max age range? If so, then it is not needed and should be set inside the function.

[lcjohnso]

I was following the format of the other prior choices. Will leave as is.

[karllark]

Nit picky point. I don't think this is a warning. There is no problem with single value parameters in a grid. I worry that having it issue a warning makes it sound like this is bad. Maybe something less "scary" like "Info"?

[lcjohnso]

Changed to "Note:"

[karllark]

Great to have this doc file! Many thanks for starting it. I know there is material I will add once I think about it more.

[coveralls]

Coverage decreased (-0.01%) to 12.153% when pulling 6e02d2d on lcjohnso:adjust_priors_lcj into 6d93d78 on BEAST-Fitting:master.