[GH-3372] Allow NaN nodata in RS_AsRaster

[mtcd001-tech]

Did you read the Contributor Guide?

Yes, I have read the Contributor Rules and Contributor Development Guide.

Is this PR related to a ticket?

Closes #3372.

What changes were proposed in this PR?

Allow NaN nodata values for float and double output in RS_AsRaster, both when supplied explicitly and when inherited from the reference raster. Preserve rejection for integer pixel types, finite-value checks for burn values, and exact storage checks for finite nodata values.

Infinity remains rejected: GeoTools cannot construct an infinite nodata category bound. This matches the existing infinite-nodata handling in RS_FromGeoTiff. The API documentation now makes this distinction explicit.

How was this patch tested?

On Windows with Java 17:

• Both new NaN regression tests fail against unchanged upstream production code and pass with the fix. They verify background pixels, nodata metadata, and the count excluding nodata for F/D output.

• Tests also cover inherited NaN rejection for integer bands, nonfinite integer nodata, infinite floating nodata, and nonfinite burn values.

• All 128 tests in RasterConstructorsTest, RasterizationTest, RasterBandAccessorsTest, RasterUtilsTest, and SerdeTest pass:

  mvn -pl common -am -Dtest=RasterConstructorsTest,RasterizationTest,RasterBandAccessorsTest,RasterUtilsTest,SerdeTest -Dsurefire.failIfNoSpecifiedTests=false test

• Including RasterBandEditorsTest gives 146 passing tests and one existing failure in testClip: a Windows line-ending difference in GeoTools text output. The same failure was reproduced against unchanged upstream production code.

• Spotless and git diff --check pass.

Spark and Python suites were not run locally.

Did this PR include necessary documentation updates?

Yes, the RS_AsRaster documentation describes support for NaN nodata on F/D and the infinity restriction.

[jiayuasu]

Thank you!