HEIF export: write HDR10 content light level (MaxCLL/MaxFALL) for PQ

[MaykThewessen]

PR: HEIF export — write HDR10 content light level (MaxCLL/MaxFALL) for PQ

Branch: upstream-heif-export (1 commit, src/imageio/format/heif.c, +58 / -0)
Base: darktable-org/darktable:master
Refs: part of #18078 (companion to the AVIF change in #21357)

---

What this does

darktable's HEIF exporter already tags output with the correct nclx colour
information (color_primaries / transfer_characteristics / matrix_coefficients)
for PQ and HLG Rec.2020 and P3 profiles, but never wrote the HDR10 content
light level box (clli: MaxCLL and MaxFALL). Players, OSes and HDR displays
use those values to tone-map, and platforms such as iOS/Instagram expect them on
HDR stills.

This adds clli for PQ output, mirroring the equivalent AVIF change.

How

For PQ (SMPTE ST 2084) the exported float samples are absolute-luminance
encoded, so the real content light levels are derived from the pixels via the PQ
EOTF:

• MaxCLL = the brightest single sample (max RGB component), in cd/m².
• MaxFALL = the mean over all samples of that per-sample peak light level.

Written via heif_image_set_content_light_level(), gated on the PQ transfer
characteristic. HLG is relative (no fixed nit scale) and is left untouched, as
is the ICC-fallback path.

Testing

Exported a 12-bit PQ Rec.2020 .heif from darktable-cli and inspected it:

heif-info:
  nclx: primaries 9 (BT.2020) | transfer 16 (PQ) | matrix 9 (BT.2020 NCL) | full range
  content light level: MaxCLL=8970 MaxFALL=4615

ffprobe:
  color_primaries=bt2020  color_transfer=smpte2084  color_space=bt2020nc
  Content light level metadata: max_content=8970 max_average=4615

Without this change the clli box is absent. (The high numbers above are from a
fully-saturated synthetic test image with no filmic roll-off; a normally graded
photo lands much lower.) Same _pq_to_nits derivation as the AVIF path, for
consistent HDR10 signalling across both formats.