Tags: DanWBR/NINA.Polaris

feat: add dust mote removal tool to Studio

- DustMoteRemoval: detect motes from star-rejected vs large-scale background,
  divide out per-channel local synthetic flat; multi-threshold detection,
  nebula-halo exclusion, star-spike pre-clip
- DustRemovalService: before/after preview + mote geometry, full-res apply to
  {stem}_dustfix.fits preserving headers + DUST* keywords
- /api/post/dust-preview and /dust-remove endpoints; register the service
- Studio modal: Sensitivity/Min size/Feather/Strength sliders, live preview,
  detected-mote overlay, hold-to-compare, opens the comparator on apply
- Strength up to 200%; responsive modal up to ~80% width
- DustRemovalServiceTests

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

fix: cap the Sharpless catalogue in Tonight's Best like LBN and LDN

- CapCatalogue(items, "Sh2", 10) next to the existing Lynds caps
- CapCatalogue is internal so the cap can be tested directly
- cap tests build SkyCatalogService over the real dso.db; with the default
  constructor it serves a legacy ~200-object list holding no Sh2 or LBN, so
  the tests counted zero and would have passed with the cap deleted
- drop a duplicated doc comment left over from the dedup change

fix: skipped dither ate the slot; overlay and Tonight's Best showed t…

…he wrong names

Live-stack dither
- a skip no longer advances the every-N gate, so the next frame retries
- the skip notice says whether the guider is disconnected or not guiding,
  and is logged once per transition instead of per frame
- the frame carries on to the recenter check after a skipped dither
- LastDitherFrame in the trigger status

Annotation overlay
- merge catalogue rows that name each other in their aliases and agree on
  position, so M27 and NGC 6853 are one label; folded names come back as `aka`

Tonight's Best
- angular-separation match instead of 2 arcsec buckets (Arp 168 and M32 are
  3 arcsec apart and both survived)
- shared DesignationRank over the catalogue column, covering Arp, HCG, UGC,
  PGC and the rest instead of tying them all at the bottom

Status bar
- host readouts after the board name fold away while activity chips are up,
  staggered, and unfold when the last chip clears; restored inside the open
  tray and under prefers-reduced-motion

- LiveStackingService.RaiseFrameIntegratedAsync extracted from AddFrameAsync
- pt-BR, es, fr, de for the new strings

Fix the three wrong answers in the plate-solve database card

- the field of view falls back to the connected camera, so a rig whose camera
  reports its pixel and sensor size no longer reads "unknown"
- the ASTAP scan only counts numbered database shards, not every underscored
  file in the folder
- ScaleOf reads index-2mass-07-03 as well as index-4209, so an installed set
  stops reporting as absent

Measured on an Orange Pi 5 Pro: fovDeg 0 with a camera reporting 2.9um and
3840x2160, ACKNOWLEDGEMENT listed as an installed ASTAP database, and 311 MB
of indexes reported as none.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Say "Start over" on the PLAN button when starting would discard progress

- planStartLabel gives three states: Stop, Start over, Start
- planStartTitle spells out how many frames a restart throws away and points
  at Resume
- four strings added to the five catalogues

The button sat next to Resume reading "Start" while it silently discarded the
retained progress.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Drop the not-tested banner from PLAN and the Advanced Sequencer

- both panels lose the caution strip; the headers stay as they were
- the class had no other users, so its styles go too, trimming the two
  grouped selectors without touching their neighbours
- stale comments that still described the banner updated
- the orphaned "Click here to report an issue" string removed from the nine
  catalogues after checking nothing else uses it

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Drop UTC from the plan start and end option labels

- "At time (UTC)" becomes "At a set time" in both selectors
- key renamed in the five catalogues and the four machine-translation maps

The option advertised UTC while the field directly below it said (local) and
converted local to UTC on save.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Anchor the plan preview at now for a Now plan, not at dusk

- openPlanViz places the cursor at Date.now() clamped to the charted night
  when startMode is Now, and keeps the AtTime branch unchanged
- dusk still drives the night shading

PlanCompilerService only emits a WaitUntilTime for AtTime, so a Now plan
starts immediately; the preview drew its window from dusk and showed an
already-finished night when the plan was built after dark.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Depend on the indi-3rdparty camera libraries instead of pretending to…

… ship them

- deb Recommends libtoupcam, libaltaircam, libasi, libsvbony, libplayerone
- drop the Linux Content items from the five native camera SDK projects

The items set CopyToOutputDirectory, which does not carry content from a
ProjectReference into the app's publish output, so no native camera library
was ever shipped. Installs worked because indi-full pulls those packages in.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Warp the CFA per plane in the mono live stack instead of warping the …

…mosaic

- BayerDebayer.Remosaic samples R/G/B planes back into a CFA at the pattern's
  phase, the inverse of the debayer's sampling step
- LiveStackingService.Warp debayers, warps each plane and remosaics when the
  session carries a Bayer pattern and colour is off
- skipped when _colorActive, where the caller warps the planes itself
- tests: lossless round trip on all four patterns, plus per-plane warp keeps
  the channels apart under a sub-pixel shift while a mosaic warp collapses them

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>