[KnifeLemon]

Part of #1027. A print user's CMYK PSD → PDF came out RGB, and a CMYK JPEG → PDF failed (fixed in #1057 by making it RGB). Print work needs its CMYK kept.

PDFsharp's image import takes grey and RGB only, so Pdf/CmykPictures.cs writes the image XObject itself for CMYK pictures:

• Colour space: the picture's own CMYK ICC profile as ICCBased (N=4, alternate DeviceCMYK), or DeviceCMYK without one. A PSD with Japan Color 2001 Coated keeps it.
• Data: a JPEG source as JPEG again (DCTDecode, the preset's quality, at least 85 as before). ImageMagick writes CMYK JPEGs inverted with Adobe's APP14 marker, as Photoshop does, so a Decode [1 0 …] array undoes it. Anything else (PSD, TIFF, …) as its CMYK samples, losslessly (FlateDecode).
• Transparency: a soft mask (SMask).
• Placement: the same page size, margins and centring as before, drawn with cm … Do in the page's content.

Grey and RGB pictures still go through PDFsharp unchanged.

Tested

New CmykPdfTests, each read back with PDFsharp and drawn back with PDFium:

• CMYK JPG, TIFF and PSD with the SWOP profile → the PDF image is ICCBased N=4 (DCT for JPG, Flate for the others), and the rendered page shows orange and navy where they are;
• without a profile → DeviceCMYK, colours right;
• the lossless samples in the PDF are byte for byte the picture's CMYK;
• a transparent half shows the white page; an RGB PNG is still DeviceRGB;
• the Adobe-marker reader.

Engines PDF/image tests 249 pass (26 skipped for engines not installed here); dotnet format --verify-no-changes clean.

The other half of #1027 (a CMYK AI / PDF → JPG keeps CMYK) comes in its own PR.

🤖 Generated with Claude Code