Fixing 955

[nagisml]

Followup of #947 fixing #955

[AgreeDK]

Thanks, @nagisml
nice, well-scoped fix. Using peek_value()/write_file() so dist_calc_* lands in the target file's own table (rather than the legacy keys while the engine is switched) fits the #659 binding rule well, and the NULL check closes a real gap in distances_up_to_date(). Tests and mypy pass here.

A few non-blocking thoughts, fine as a follow-up:

While the worker recalculates, the progress bar sits at 100% with no feedback. A self.progress.emit(0, 0) just before recalculate_distances() would switch it to busy mode (the dialog already handles total == 0).
When several GSAK databases go into the same target, each job does a full recalc. That's correct, just redundant; recalculating once after the last job per target could be a later optimisation.
The new NULL check means an up-to-date database pays for a full table scan on startup/switch (no index on distance). Probably fine, but worth a quick check on a large DB.

Approving — thanks again!