Fix outliers() crash on bounded parameters with string dimension limits

[arpitjain099]

Hi, thanks for maintaining pyPRMS.

Parameter.outliers() raises for any bounded parameter. The metadata loader keeps a dimension name as the maximum (for example nlake on lake_hru_id) when it can't cast the bound to the parameter datatype, so outliers() ends up doing a numpy > between the int data and a string and hits a _UFuncNoLoopError. It affects the whole set of bounded params (lake_hru_id, hru_solsta, the cascade _down_id/_up_id parameters, and more).

check_values() already skips the comparison when a bound is a string, so this just brings outliers() in line with that. Numeric bounds behave exactly as before; string bounds are skipped and no longer counted.

Added a case to test_Parameter covering the bounded parameter.

[pnorton-usgs]

Hi Arpit,

Thank you for your PR on the outliers() crash with bounded parameters. As it turns out this was indication of a greater problem in the code. The intent for bounded parameters was that the minimum and maximum metadata values would be changed to integer values when a bounded parameter was added reflecting the full range of acceptable values for the parameter. In practice this was not happening in all cases. This problem was addressed in PR #63 and #65 so bounded parameters should now be fully resolved when they are added and the outliers() function should work properly with them. Currently poi_gage_id is the only true string parameter where the minimum and maximum metadata values would be strings; so when a parameter is a string the outliers() code will return an empty list. Non-string parameters should raise an error if minimum or maximum metadata is a string.

After applying fixes the changes you have suggested no longer apply. Thank you for your help in uncovering this problem. I will close this PR at this time.