Sync: Add contact-table sync via cloud-folder journals

[aa5sh]

Multiple QLog installs that share a cloud folder (Dropbox, iCloud Drive, OneDrive, Syncthing, ...) keep their contacts tables in sync. Each install writes its own append-only JSON-lines journal into /qlog-sync/nodes//journal-NNNN.jsonl and reads every other node's journal back, applying remote upserts and deletes locally with last-writer-wins by updated_at.

Database (migration 039)

• New columns on contacts: qso_uuid (UUIDv4), updated_at (ISO-8601 UTC), origin_node.
• New contacts_tombstones(qso_uuid PK, deleted_at, origin_node) so hard deletes are observable for sync without changing any existing SELECT path.
• New sync_runtime(key, value) so the AFTER INSERT / AFTER UPDATE / BEFORE DELETE triggers can read the current node id at fire time; the user can rename the node from the Sync dialog without rebuilding any trigger.
• New sync_peers(node, last_file, last_offset, last_seen_at) for per-peer reader resume.
• AFTER INSERT trigger auto-fills qso_uuid / updated_at / origin_node when a write path didn't set them, so manual log entry, ADIF import, WSJT-X and fldigi all get sync metadata for free.
• AFTER UPDATE trigger bumps updated_at on every local edit and re- stamps origin_node to the current self_node_id, so an edit to a row that originally came from a peer is "claimed" by the editor and shows up in the writer's "origin_node = self" filter.
• BEFORE DELETE trigger records a tombstone.
• Migration also backfills existing rows with UUIDs, sensible timestamps, and the device hostname as the default node id.

ContactSync (core/ContactSync.{h,cpp})

• Singleton with a 30 s timer that runs a flush + pull cycle, plus a requestFlush() slot wired to contactAdded / contactUpdated / contactDeleted so local UI changes hit the journal in the next event-loop tick (deferred via QTimer::singleShot to fire after QSqlTableModel::beforeUpdate / beforeDelete commits).
• flushOnDb writes new local upserts and tombstones into the current journal file, rotates at 4 MB, and maintains a HEAD pointer + one-time manifest.json.
• pullOnDb scans sibling nodes/* directories, follows each peer's journal-*.jsonl files in name order past sync_peers.last_offset, and applies each record.
• applyUpsert resolves identity by qso_uuid: LWW if matched, honour tombstones, otherwise fingerprint-match (callsign + mode + band + sat_name + start_time within ±30 min, same fields as ADIF import) and either Skip or Merge by the user-chosen DuplicatePolicy. Merge converges peers on the lexicographically- smaller UUID — both peers reach the same decision without coordination, then LWW reconciles fields.
• applyDelete deletes by qso_uuid (LWW vs local updates), or records a tombstone when no local row exists so a subsequent older insert from another peer is filtered out.
• Internal *OnDb methods take QSqlDatabase &db so a future move to a worker thread can pass a per-call cloned connection without changing call sites. The dispatchers currently run synchronously on the main thread.
• snapshotForCycle() caches folder + self for the duration of a cycle.
• busyChanged(bool) signal lets the dialog gate its buttons.
• LogParam grows typed wrappers for the seven sync settings.

UI (ui/SyncDialog.{h,cpp,ui})

• New dialog under Logbook → "Contact Sync…" with an enable toggle, folder picker, editable Node ID + Save, duplicate-policy combo, last-flush + last-pull timestamps, pending-changes counter, and Flush Now / Pull Now buttons.

MainWindow wiring

• contactAdded / contactUpdated / contactDeleted signals fire ContactSync::requestFlush via QTimer::singleShot(0); using a direct connection because Qt's queued connections can't copy a QSqlRecord& reference argument.
• ContactSync::pulled (when applied > 0) triggers LogbookWidget::updateTable so remote-sourced rows appear in the logbook view without manual refresh.
• ContactSync::start() runs on app launch when sync is enabled.