Make acquisition scientifically reliable end-to-end

[edjuh]

Current gaps to track after the 2026-05-25 run:

• Science frames must not start unless Alpaca tracking is verified ON after slew and before exposure.
• Trailed frames/stacks must be rejected before photometry/AAVSO acceptance.
• Postflight should publish one solved/accepted stacked preview per object per night, not raw-frame preview spam.
• Need tighter proof that telescope is synced/on-target after slew: solve, nudge/sync, then expose.
• Stacking should be alignment-aware and must fail clearly when registration quality is poor.
• Dashboard/logs should expose tracking state, solve status, stack status, and reject reason per object.

Local/RPi hardening started with strict tracking and star-shape QC gates; this issue remains the umbrella for the remaining reliability work.