[rnissl]

Summary

• Treat GPS time as usable only after GPS9 day 8000 and a 10/100 ms-aligned clock (Hero 11 / Max 2 indoor default calendar).
• CAMM extraction can stop at the first good point.

Test plan

• existing CAMM/geo tests still pass

Rebased onto current mapillary/mapillary_tools main (d1b82ab). Independent of the GoPro Max / custom FFmpeg stack.
CAMM extract conflict with #828 resolved by keeping epoch normalization and stopping at the first usable GPS clock.

[caglarpir]

The default dates the HERO11 and MAX 2 write before they have a real GPS time are a nasty one, good catch. I can't merge it as is, though: nothing calls gps_datetime_is_valid or first_gps_only yet, so behaviour doesn't change, and GPS9 comes from GoPro's GPMF track while this changes the CAMM parser, so I'm not sure which videos it would apply to. 0.15.0 also reworked how a video's start time is read from its GPS clock (#832, #939). If you still get wrong times with 0.15.0, could you open an issue with a short clip that shows the problem? With a sample we can find a check that's more robust than a day cutoff. Closing this one in the meantime.