bronx-19 fregrid is broken for stretched grid sources

[nikizadehgfdl]

bronx-19 fregrid is broken for stretched grid sources

[nikizadehgfdl]

@ceblanton or @ngs333 can you please assign this issue to me since I am working on it.

[nikizadehgfdl]

I found the root cause. The fix I made for poly_area in issue #45 produces the corrected value for area in grid cells that straddle the South Pole that would otherwise be 25% of the area of the Earth.
But, the function poly_ctrlat() in create_xgrid.c which is called by frergid --conserve_order2 has a similar bug for those grid cells and comes up with a vastly wrong latitude for the centroid of those grid cells. However since ctrlat is devided by cell_area in fregrid calculations the two nonsense huge values (for incorrect area and ctrlat) used to cancel each other out and fregrid would end up with a finite value.
After my fix to poly_area the cancellation would not happen and we end up with very wrong (huge negative) values for otherwise positive quantities (e.g., olr) for the few grid cells around the South Pole.
The fix is simply to do the same fix I did for poly_area() in poly_ctrlat().

In my opinion the fregrid estimates are not accurate at the South Pole cells for stretched grids, with or without any of my fixes. It's only that the values make sense after I put the fix in both poly_area and ctrlat, or without either of the fixes (as in bronx-18).