Merge PR #699: LR-PROF-CAMERACOLOR: fit the tone curve of low-contrast scenes (range limit 1.5 to 1.2)

crates/engine/src/camera_preview.rs: LOOK_VERSION keeps both history lines and the higher value.
crates/engine/src/media.rs: RENDER_CACHE_VERSION keeps the higher value.

docs/camera-preview-colour.md

@@ -4,8 +4,7 @@ ARW, NEF, RW2, RAF, CR3, CR2, PEF, SRW, ORF and ERF decoding supply sensor data
 
 ## Colour and tone are separate
 
-A fixed sensor proxy of at most 96×96 is developed with the file's WB, black/white levels and active crop. The embedded JPEG is decoded with DCT scaling and a 64-million-pixel input limit, converted through the codec's colour management, and reduced to the same size. Aspect ratios must agree within 2%. For colour fitting, near-black, clipped and nonfinite samples are excluded. Learning a colour matrix requires at least 256 paired pixels, 5% coloured samples (linear channel spread >0.05) and a nonsingular input covariance with bounded condition. An existing camera profile supplies the colour model: fitting only tone still requires 256 pairs and 5% non-monochrome samples, but permits channel spread >0.005 so dark or dull scenes can use the known colours.
-A fixed sensor proxy of at most 96×96 is developed with the file's WB, black/white levels and active crop. The embedded JPEG is decoded with DCT scaling and a 128-million-pixel input limit (61 MP bodies' JPEGs, e.g. the ILCE-7RM6 full frame at 66.5 MP, exceed 64 MP; the codec checks the cap before decoding at full size, about 200 MB transient), converted through the codec's colour management, and reduced to the same size. Aspect ratios must agree within 2%. Near-black, clipped and nonfinite samples are excluded. Learning a colour matrix requires at least 256 paired pixels, 5% coloured samples (linear channel spread >0.05) and a nonsingular input covariance with bounded condition. An existing camera profile supplies the colour model: fitting only tone still requires 256 pairs and 5% non-monochrome samples, but permits channel spread >0.005 so dark or dull scenes can use the known colours.
+A fixed sensor proxy of at most 96×96 is developed with the file's WB, black/white levels and active crop. The embedded JPEG is decoded with DCT scaling and a 128-million-pixel input limit (61 MP bodies' JPEGs, e.g. the ILCE-7RM6 full frame at 66.5 MP, exceed 64 MP; the codec checks the cap before decoding at full size, about 200 MB transient), converted through the codec's colour management, and reduced to the same size. Aspect ratios must agree within 2%. For colour fitting, near-black, clipped and nonfinite samples are excluded. Learning a colour matrix requires at least 256 paired pixels, 5% coloured samples (linear channel spread >0.05) and a nonsingular input covariance with bounded condition. An existing camera profile supplies the colour model: fitting only tone still requires 256 pairs and 5% non-monochrome samples, but permits channel spread >0.005 so dark or dull scenes can use the known colours. Any tone curve (full, partial or black-and-white look) also needs some contrast: the scene luminance of the brightest 1/32 of the pairs must be at least 1.2× that of the darkest 1/32 (`MIN_TONE_RANGE`); below that, the range check refuses the curve, which leaves only nearly uniform frames. A curve under 1.5× (the limit until issue #631) must also render the frame's bright areas, which the fit and the held-out gates leave out (camera JPEG luminance 0.85 and up), no further from the camera JPEG than the neutral fallback does (in luminance for a black-and-white JPEG): fitted to a narrow range, it extrapolates them. A hazy ILCE-7RM4 burst at 1.46 to 1.49 opened with the neutral fallback, 18 L\* darker than its camera JPEG, while the frame at 1.51 got a look at a held-out RMS of 0.013; with 1.2 they all do. Across 135 public Sony files the lowest range is 4.0, so none of them changes.
 
 Two thirds of the usable pixels train a ridge-regularised 3×3 **chromaticity** correction (luminance-normalised RGB). This avoids trying to encode the camera's nonlinear tone curve in a matrix, which left skin grey in the earlier matrix-only estimate. Coefficients are bounded by 8. A matrix can't follow the camera's hue-dependent rendering, though: on a photo of a lime shirt among grey rocks the best matrix rendered the shirt olive (16° hue error), because the few saturated pixels have little weight against the many dull ones, and a matrix fitted to the shirt alone breaks everything else. So a **hue/saturation table** (5 values × 72 hues × 5 saturations, the DNG `ProfileHueSatMap` layout with an sRGB-encoded value axis, applied in linear ProPhoto RGB after the baseline exposure) is fitted to the residual hue shifts and saturation ratios after the matrix: Gaussian kernels around each node (2.5° hue, 0.15 saturation, 0.12 encoded value), weighted by saturation, shrunk toward identity where the photo has few samples, saturation 0 left as identity so neutrals stay neutral. The value axis matters: on a sunny waterfall scene dark yellows had to turn ~12° toward orange while brighter yellow-greens turned ~9° toward green, and one shift per hue made both worse. The table restores each pixel's luminance (the tone curve owns luminance; without that, saturation boosts darkened colours and the table lost on held-out pixels).