Data requests for legacy datasets fail

Service error:

java.lang.RuntimeException: Bad ranges [] length
	at gov.usgs.earthquake.nshmp.netcdf.reader.BoundingReaderHazardCurves.mapDataFromArray(BoundingReaderHazardCurves.java:150)
	at gov.usgs.earthquake.nshmp.netcdf.reader.BoundingReaderHazardCurves.extractDataAt(BoundingReaderHazardCurves.java:93)
	at gov.usgs.earthquake.nshmp.netcdf.reader.BoundingReaderHazardCurves.setBoundingData(BoundingReaderHazardCurves.java:172)
	at gov.usgs.earthquake.nshmp.netcdf.reader.BoundingReader.<init>(BoundingReader.java:33)
	at gov.usgs.earthquake.nshmp.netcdf.reader.BoundingReaderHazardCurves.<init>(BoundingReaderHazardCurves.java:30)
	at gov.usgs.earthquake.nshmp.netcdf.NetcdfHazardCurves.boundingData(NetcdfHazardCurves.java:32)
	at gov.usgs.earthquake.nshmp.netcdf.NetcdfHazardCurves.staticData(NetcdfHazardCurves.java:42)
	at gov.usgs.earthquake.nshmp.netcdf.www.NetcdfServiceHazardCurves.processCurves(NetcdfServiceHazardCurves.java:99)
	at gov.usgs.earthquake.nshmp.netcdf.www.NetcdfServiceHazardCurves.processRequest(NetcdfServiceHazardCurves.java:158)
	at gov.usgs.earthquake.nshmp.netcdf.www.NetcdfServiceHazardCurves.processRequest(NetcdfServiceHazardCurves.java:34)

See ghsc/nshmp/nshmp-netcdf-conversion#31

Possible solutions:

1. Prevent the singleton dimension from being dropped when the netCDF file is read in
2. Only add the site class index to the shape and origin vectors if the number of siteClasses is greater than one