feat(macos): EDR and HDR PQ/HLG presentation modes for the Metal view

[mcoliver]

Summary

Adds EDR and HDR (PQ/HLG) presentation to the macOS Metal view. SDR stays the default.

What was wrong

• Everything above SDR white was clipped. The Metal path renders into an RGBA16F FBO, then copies each frame into a 10-bit unorm IOSurface. So the output was 10-bit SDR, even though the layer sets wantsExtendedDynamicRangeContent and the README claimed EDR.
• The IOSurface colorspace tag was ignored. IOSurfaceSetValue(surf, kCVImageBufferCGColorSpaceKey, cs) stores a private CGColorSpace key, not the IOSurfaceColorSpace attachment Core Animation reads.

Changes

• MetalPresentationFormat (new, Qt-free): defines each mode (IOSurface pixel format, matching GL formats, colorspace) and creates IOSurfaces tagged through CoreVideo (CVBufferSetAttachment), which writes IOSurfaceColorSpace/IOSurfaceColorSpaceID.

  Mode	IOSurface	Tag	Use
  sdr (default)	ARGB2101010	sRGB	unchanged behaviour
  edr	64RGBAHalf	extended sRGB	IPCore's unclamped sRGB-encoded output passes through; values above 1.0 use the display's EDR headroom
  pq	ARGB2101010	BT.2100 PQ	display colorspace SMPTE 2084 or an OCIO PQ view
  hlg	ARGB2101010	BT.2100 HLG	display colorspace HLG or an OCIO HLG view

• QTMetalVideoDevice / MetalView: the zero-copy ring, the CPU fallback and the fallback upload now use the selected format.

• Choosing a mode: it's read once at startup from UTV_MACOS_PRESENTATION or Preferences → Rendering → macOS Presentation (Display/macosPresentation). The control is added in code (macOS only), so the shared .ui files don't change.

• Logging: when an HDR mode is active, the display's current/potential EDR headroom is logged.

• README: now describes the macOS modes accurately. docs/testing/macos-edr.md has test instructions.

Verification (Apple Silicon, macOS 27)

• New MetalPresentationTest (headless, doctest, 41 assertions). For each mode it checks the IOSurface pixel format and colorspace tag. It then runs the same GL path as the zero-copy ring (RGBA16F FBO → blit → IOSurface-backed texture) and reads the IOSurface back:
  • EDR keeps 4.0.
  • SDR clips to 1.0.
  • PQ/HLG keep 10-bit precision.
• IPCore doesn't clip values above 1. utvio renders linear 4.0 with -outsrgb as 1.825, exactly the extended-sRGB value.
• This MacBook's built-in display reports potential EDR headroom 2.0×, so EDR can be checked visually without an HDR monitor.
• ctest: 119/119 pass.

Not yet verified

• Visual check on a display with EDR headroom, using the three-tile 1×/2×/4× test in the doc. I couldn't do this: my session had no attached display.
• PQ/HLG with real HDR content and an OCIO HDR view.
• The new test runs in CI once test(ci): add CLI smoke and image format round-trip tests; run ctest in CI #64's ctest step lands. On GPU-less runners it falls back to the software renderer and skips the GL round trip if IOSurface binding is unavailable.

Follow-ups

• Tone-map or roll off highlights to the reported headroom, instead of letting macOS clip, and show headroom in the HUD.
• React to headroom changes (display switches, brightness) via NSScreen notifications.
• A cross-platform presenter on Qt's QRhi swapchain (HDRExtendedDisplayP3Linear / scRGB / HDR10) behind the same format abstraction.

🤖 Generated with Claude Code