[GH-3245] Fix RS_AsCOG ArrayIndexOutOfBoundsException for byte-band rasters

[jiayuasu]

Did you read the Contributor Guide?

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

Is this PR related to a ticket?

• Yes, and the PR name follows the format [GH-XXX] my subject. Closes RS_AsCOG throws ArrayIndexOutOfBoundsException for 1024x1024 byte-band rasters #3245

What changes were proposed in this PR?

RS_AsCOG threw ArrayIndexOutOfBoundsException at Deflater.setInput for 1024x1024 byte-band rasters. The root cause is in imageio-ext's TIFFDeflater: it passes (offset, height * scanlineStride) verbatim to Deflater.setInput(). The COG pipeline's overview images are JAI PlanarImages tiled wider than the 256px output tile (a 1024 input produces a single-tile 512x512 overview); TIFFImageWriter's 8-bit optimized path hands the compressor the overview's backing array with a tile offset and the full scanline stride, and for the last row of output tiles the computed range overruns the array. Only byte bands take that direct-buffer path, which is why int/double rasters were unaffected. The bug is still present in imageio-ext 2.1.0.

Two related consequences of the same code path are also covered:

• When it did not crash, it compressed the inter-row slack between tile width and scanline stride, silently corrupting overview pixels for non-empty data.
• The full-resolution write crashes the same way for any byte raster whose image is tiled wider than the output tile size (e.g. loaded from a 512-tiled GeoTIFF).

The fix retiles byte-band coverages to the output tile grid in CogWriter before handing them to GeoTiffWriter, so every raster the writer reads starts at offset 0 of a tile-sized buffer. The copy happens lazily one tile at a time via javax.media.jai.TiledImage; pixel data, geo-referencing, sample dimensions, and metadata are unchanged, and non-byte bands are untouched.

How was this patch tested?

New regression tests in CogWriterTest:

• testWriteByteBandRasterAsCog: byte rasters at 512/768/1024/1536 write successfully and full-resolution pixel values round-trip exactly (the 1024 and 1536 cases crashed before this fix).
• testByteBandOverviewPixelsSurviveCogEncoding: reads the overview IFD back and verifies overview pixel content, catching the silent stride corruption.
• testWriteTiledByteSourceAsCog: a byte coverage whose own image is 512-tiled writes correctly with overviewCount=0, covering the full-resolution-path variant.

Full common module suite passes (1288 tests).

Did this PR include necessary documentation updates?

• No, this PR does not affect any public API so no need to change the documentation.