Unify daily dog-work status

[OwenTanzer]

Summary

• derive one canonical per-dog session count from the report sessionDate and local calendar date
• show neutral, worked-once, and distinct worked-multiple-times states on dog cards, profiles, pinned dogs, and dashboard activity lists
• limit Needs Attention to currently assigned dogs in the pinned folder with no session yesterday
• exclude released, graduated, and transferred source dogs from Recently Worked
• roll all derived states over automatically at local midnight

Reset behavior

The indicator remains derived-only rather than adding a manual reset override. A resettable flag could disagree with the actual logs and with backdated edits; deleting, correcting, or backdating the underlying report immediately recomputes the canonical count instead.

Validation

• npm test (20/20)
• npm run build
• npx tsc -b
• npm run lint (only three pre-existing warnings)
• npm --prefix worker run typecheck
• git diff --check

Closes #47

[OwenTanzer]

daily counts are memoized against an array that is mutated in place

useDailySessionCounts() memoizes against [reports, today].

However, the store deliberately mutates nested arrays and objects in place, then only shallow-clones the outer db object to trigger rendering. In particular:

createReport() uses db.reports.push(report).
updateReport() mutates the existing report with Object.assign.
Only deletion replaces the reports array with a new array.

React therefore rerenders after an edited report, but reports still has the same reference, so the memoized count is reused. On the dog profile:

changing a log from yesterday to today may leave the dog showing not worked today;
changing today’s log to another date may leave Worked once today visible;
changing dates among multiple sessions may leave the wrong once-versus-multiple state.

That directly contradicts the PR’s stated derived-only behavior, where correcting or backdating a report should immediately recompute the canonical count.

Fix: depend on the full database snapshot rather than the nested array reference, or simply compute the inexpensive count without useMemo, for example:

export function useDailySessionCounts(): Record<string, number> {
const state = useDatabase();
const today = useCurrentLocalDate();

return useMemo(
() => sessionCountsByDogOnDate(state.reports, today),
[state, today],
);
}

A component-level regression test should simulate an in-place sessionDate edit followed by a new outer-store snapshot. The current tests validate the pure counting and assignment functions but cannot expose this React memoization failure.

Everything else is well aligned with #47: zero/one/multiple states are distinct, attention is limited to assigned dogs in the pinned folder, transferred sources are excluded, and midnight rollover is explicitly handled. CI is fully green.

[OwenTanzer]

Addressed in 4c6e637. I removed the memoization entirely, so useDailySessionCounts() now derives the inexpensive count from the current report contents on every database render and cannot be held stale by a stable nested-array reference. I also added a regression that keeps the same reports array/object, mutates sessionDate in place, appends in place, then backdates in place, verifying the counts transition 0 → 1 → 2 → 1. Local validation passes: 21 tests, app and worker typechecks, lint (three pre-existing warnings only), diff check, and production build.