Enhance Point-Stat and Ensemble-Stat to support new masking options to filter point observation locations directly.

[JohnHalleyGotway]

This request for enhancement came via MET-Help with Bob Craig at the Air Force:
https://rt.rap.ucar.edu/rt/Ticket/Display.html?id=91234

Suggest adding 2 new options to the "mask" dictionary, "gridpnt" and "polypnt".

The mask section in Point-Stat currently contains 4 entries for defining masking regions. But there is a subtle difference between them:

• The grid and poly options are "area" masks, meaning they define which FORECAST grid points should be included in the verification region. Any forecast grid points outside those regions are considered to contain bad data values.
• The sid and llpnt entries are "point" masks, meaning they define which point OBSERVATIONS should be used. The mask.sid option specifies the list of station names while the mask.llpnt option defines thresholds for the lat/lon values (with no consideration given to the grid on which the vx is done).

These sort of details matter a lot when dealing with edge effects. The appeal of the mask.grid and mask.poly options is that the same logic for them is applied consistently throughout the MET tools (in Point-Stat and Grid-Stat, in particular). But perhaps there's room to add more masking options which would apply the grid and poly masking logic to the point OBSERVATION locations instead of the forecast grid points.

For example, "gridpnt" and "polypnt" could mean "use point observations falling inside these regions". And don't apply them in any way to the gridded forecast data. I think this logic would make it easier to compare the performance of the nest to the parent using the same set of point observations... basically solving the edge effect issue.

Feedback from Bob Craig on 07/26/2019:
I think it would be a very help full addition. We do comparisons of nested models all the time an want to have the same obs used for the computation of statistics.