Add exhaustive NAD83 -> NAD83(HPGN) shift grids

[rouault]

The main proj-datumgrid package contains 5 grids (FL, MD, TN, WI, WO)
out of the full 51 grids for the 51 states. So we provide them in an
extensive way by using the build_nad83_harn.sh script that downloads
the source .los/.las files, and use the scripts/loslas2ntv2.py script
to convert them to NTv2.

Important note: https://download.osgeo.org/proj/hpgn_ntv2.zip contains
a similar content, except that this historical package had the sign of
the offsets reverse (so doing NAD83(HPGN) -> NAD83 conversion), because
of the requirements of the historical +nadgrids, that required grids to
go from a source CRS to WGS84 (here NAD83 was considered as being WGS84)
With the new +proj=hgridshift operator, we don't need such hacks, and
it is better to keep the grids as close as possible as their sources.

Discussion:

1. should we keep the 5 grids (FL, MD, TN, WI, WO) that have equivalent content to flhpgn.gsb, mdhpgn.gsb, etc... ?
2. the size of the 51 grids is 1.5 MB. Should they be kept in north-america, or are they general purpose enough to be moved to main ?

[mwtoews]

I'd say the quantity and limited geographic extent of these files suggests they be kept in north-america

[mwtoews]

I wonder if a metadata table would help identify the filenames and regions. See (e.g.) geographic_transformations.pdf for these relations, but perhaps a .csv file. For instance, it's not obvious that c1hpgn.gsb is for Saipan, near Guam, in the Pacific Ocean. And yeah, not all of these are in North America (i.e. not part of NAD).

[rouault]

I wonder if a metadata table would help identify the filenames and regions

The metadata table will be the EPSG database (https://github.com/rouault/proj.4/blob/iso19111/data/sql/grid_transformation.sql) + the https://github.com/rouault/proj.4/blob/iso19111/data/sql/grid_alternatives.sql (to be completed to do the mapping between the .los/.las names and the .gsb ones) as part of my work of https://github.com/rouault/proj.4/tree/iso19111

[kbevers]

Blah.. I was hoping to not have to make up my mind about this :-)

should we keep the 5 grids (FL, MD, TN, WI, WO) that have equivalent content to flhpgn.gsb, mdhpgn.gsb, etc... ?

I guess these have been in circulation long enough that we better keep them around. At least for now.

the size of the 51 grids is 1.5 MB. Should they be kept in north-america, or are they general purpose enough to be moved to main ?

Let's put them in north-america. I think we should try not put too much stuff into the main package.