grib2: consolidate scaled value decoding

[cambecc]

Description of Changes

The code for decoding GRIB2 scaled values is duplicated in a few places. The decoding process also accumulates more floating point rounding error than it needs to.

This change:

• introduces a utility method (and tests) for scaled value decoding
• changes all locations that do scaled value decoding in Grib2Gds and Grib2Pds to use this utility method
• reduces floating point rounding error when decoding

PR Checklist

• Link to any issues that the PR addresses
• Add labels
• Open as a draft PR
  until ready for review
• Make sure GitHub tests pass
• Mark PR as "Ready for Review"

[lesserwhirls]

This looks great - thank you for the cleanup! We do have one new failing test in our extended test suite. In cdm-test/src/test/java/ucar/nc2/iosp/grib/TestGribSpheroids.java:

netcdf-java/cdm-test/src/test/java/ucar/nc2/iosp/grib/TestGribSpheroids.java

Lines 105 to 112 in 9a528fc

	public void code7_oblate_specified_m() throws IOException {
	String filename = dir + "TT_FC_INCA.grb2";
	try (NetcdfFile ncfile = NetcdfFiles.open(filename, null)) {
	Variable v = ncfile.findVariable("LambertConformal_Projection");
	Attribute axis = v.findAttribute("semi_major_axis");
	Assert.assertEquals(6377397., axis.getNumericValue().doubleValue(), 0.1);
	}
	}

the assert now throws:

java.lang.AssertionError: expected:<6377397.0> but was:<6377397.16>. I think this is easily explained by the reduced floating point rounding error when decoding. If you agree that is the case, would you mind updating the check in the test. I've attached the test file (that is not available in the codebase) in case you are interested in the specific case (compressed with zip so that I can attach it to the issue).

TT_FC_INCA.grb2.zip

[cambecc]

The change from 6377397.0 to 6377397.16 is due to widening the Grib2Gds earthRadius, majorAxis, and minorAxis fields from float to double (and also using double for scaled value decoding). I checked the grib octets and they have the majorAxis as 637739716 with a scale factor of 2. So I think this increase in precision is correct. I will update the test.

[lesserwhirls]

Looks great - thank you!