updates for new netcdf files including PGV

• bumped nshmp-lib version (for NerhpSiteClass)
• update resources
• update code for new netcdf file structure - See ghsc/users/jaltekruse/netcdf-test#1
  • imt and siteClass enums (nshmp-lib enums Imt and NehrpSiteClass)
  • iml variable has dimensions [imt,iml] (no longer dependent on siteClass
  • hazards variable has dimensions [siteClass,imt,lat,lon,iml]
  • variables siteClass and vs30 define mapping of these parameters
  • calculation bounds are stored as a GeoJson string
  • add file and variable metadata within netCDF

PGV data is currently preliminary, having been calculated without basin effects.

nshmp-netcdf does not yet:

• use metadata
• allow backwards compatibility in file format
• use the new gridMask netcdf variable to check for data
  • note that the PGV calculation includes slightly fewer site grid points than the COUS 2018a calculation, likely due to slightly different geojson site regions(?) (e.g. run ncdump -v gridMask nshmp_2018a_v1.1-CONUS-hazards-with-preliminary-PGV-fv0.3.nc | less). The non-zero values in gridMask indicate the number of hazard curves written for a given grid point (8 site classes * 23 Imts = 184), a handful of sites (76 out of 358465) have fewer hazard curves.

Resolves ghsc/nshmp/nshmp-netcdf#14

Interestingly, there are 76 grid points with fewer than the expected 184 hazard curves (8 site classes * 23 Imts):

$ ncdump -v gridMask ../data/nshmp_2018a_v1.1-CONUS-hazards-with-preliminary-PGV-fv0.3.nc
...
       gridMask =
  _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, 
          _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, _, 
...
          _, 184, 184, 184, 184, 184, 184, 184, 184, 184, 184, 184, 184, 184, 
          184, 184, 184, 184, 184, 184, 184, 184, 184, 184, 184, 184, 184, 
          184, 184, 184, 184, 184, 184, 184, 184, 184, 184, _, _, _, _, _, _, 
...
$ ncdump -v gridMask nshmp_2018a_v1.1-CONUS-hazards-with-preliminary-PGV-fv0.3.nc | sed -e '1,/gridMask =/d' -e 's/^ *//' -e 's/, */,/g'  -e 's/;//' | perl -n -a -F, -e '/^\s*$/ && next;  /^\}/ && next; chomp; $counts{$_} += 1 foreach grep /\d/, @F; END{ print("$_: $counts{$_}\n") for sort {$a<=>$b} keys(%counts); }'
88: 1
162: 1
176: 74
184: 358389

The sites with 176 hazard curves were calculated in the 2018a calculations but not for PGV using the map-conus.geojson file. The two sites with even fewer hazard curves (88 or 162), may be missing data in the 2018a files (TBD).

mentioned in commit 26d53392

merged