Drop usage of GDALPolygonize in DT_DumpAsPolygons

[komzpa]

​http://brian-mcgill-4.ums.maine.edu/postgis_zonal.pdf :

However this function is
inadequate for the current purpose – it merges pixels with the same value into larger polygons. In
principle this should still be acceptable but the actual implementation (in a poorly documented
“feature”) does this by converting every value to an integer. So pixels with a temperature of
13.49 and 12.51 are merged incorrectly (and the value stored is 13 which is also incorrect) 

Looking at code we have branching on whether GDAL>F<Polygonize is available, and that looks hardcoded to "no". We're supporting GDAL 1.10+ and GDALFpolygonize is available since 1.9.0, so branching can be removed.