Implement Duplicate Registration Detection in Aggregation

Implement the duplicate detection as a pure Dart function `List detectDuplicates(List activities)` — no Supabase calls, no side effects. This makes it trivially testable. The algorithm: (1) filter out activities with null contact_id; (2) sort by (peer_mentor_id, contact_id, activity_date, start_time); (3) iterate with a sliding window comparing adjacent records that share the same (peer_mentor_id, contact_id, date) key; (4) check time overlap with the condition `a.start_time < b.end_time && b.start_time < a.end_time`. Wrap the call to this function in a try/catch in the aggregation service and fall back to an empty list on any error — this ensures the non-blocking guarantee is enforced at the call site.

For the overlap_description string, format it as `'{aStart}–{aEnd} overlaps with {bStart}–{bEnd}'` using 24-hour HH:mm notation for consistency with Norwegian conventions. Add a TODO comment for a future enhancement: push this detection to a SQL window function once the activity volume at national scope warrants it.

Unit tests in Dart (flutter_test): test the duplicate detection algorithm in isolation by passing a list of mock `Activity` objects and asserting the returned list of `BufdirDuplicateFlag`. Test cases must include: (1) empty input → empty flags; (2) two records, same peer mentor + contact + date, overlapping times → one flag with correct record IDs and overlap description; (3) two records, same peer mentor + contact + date, non-overlapping times (e.g., 09:00–10:00 and 11:00–12:00) → zero flags; (4) null contact_id → excluded from detection, zero flags; (5) three records A/B/C where A overlaps B and B overlaps C → two flags (A-B and B-C); (6) duplicate detection throws an internal exception → caught, empty list returned, error logged. Integration test: verify the full `aggregateActivities` call returns a payload with populated `duplicate_warnings` when seeded data contains known duplicates.