fix: deployment capture coords now reach existing locations

The /deploy classify "Assign to existing location" path dropped the captured
GPS — only "Create new location" applied it — so units assigned to pre-existing
coordless locations left those locations without a pin.

- Classify (promote) now backfills the captured GPS onto an existing location
  that has no coordinates (doesn't clobber operator-set coords).
- Add "Reforward info" button on Assigned deployment cards + endpoint
  POST /pending/{id}/resync-location that re-pushes a capture's GPS onto its
  assigned location (explicit action, overwrites). Fixes already-classified
  locations and guards against this recurring. Logged to unit history.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

backend/routers/pending_deployments.py

@@ -296,6 +296,12 @@ async def promote_pending(
         if not location:
             raise HTTPException(status_code=404, detail=f"Location {location_id!r} not found.")
         project_id = location.project_id
+        # Backfill the captured GPS onto the existing location if it doesn't
+        # have coordinates yet. (Previously the captured coords were dropped on
+        # the assign-to-existing path, so only create-new locations got a pin.)
+        # Don't clobber coordinates an operator already set.
+        if pd.coordinates and not (location.coordinates or "").strip():
+            location.coordinates = pd.coordinates
     else:
         # Create-new path.  Need a project (existing or new).
         project_id = payload.get("project_id")
@@ -456,6 +462,50 @@ async def cancel_pending(
     return {"success": True, "cancelled_at": pd.cancelled_at.isoformat()}
 
 
+@router.post("/pending/{pending_id}/resync-location")
+async def resync_location(pending_id: str, db: Session = Depends(get_db)):
+    """Re-push a promoted capture's GPS onto its assigned location.
+
+    Use when a capture's coordinates didn't land on the location (e.g. it was
+    assigned to a pre-existing location that had none). Unlike the auto-backfill
+    on classify, this is an explicit operator action and OVERWRITES the
+    location's coordinates with the captured GPS.
+    """
+    pd = db.query(PendingDeployment).filter_by(id=pending_id).first()
+    if not pd:
+        raise HTTPException(status_code=404, detail="Pending deployment not found.")
+    if pd.status != "assigned" or not pd.resulting_assignment_id:
+        raise HTTPException(status_code=400, detail="Only a promoted (assigned) capture can be re-forwarded.")
+    if not (pd.coordinates or "").strip():
+        raise HTTPException(status_code=400, detail="This capture has no GPS coordinates to forward.")
+
+    asg = db.query(UnitAssignment).filter_by(id=pd.resulting_assignment_id).first()
+    if not asg:
+        raise HTTPException(status_code=404, detail="Resulting assignment not found.")
+    location = db.query(MonitoringLocation).filter_by(id=asg.location_id).first()
+    if not location:
+        raise HTTPException(status_code=404, detail="Assigned location not found.")
+
+    old = location.coordinates
+    location.coordinates = pd.coordinates.strip()
+
+    _record_history(
+        db, unit_id=pd.unit_id,
+        change_type="deployment_coords_reforwarded",
+        old_value=old,
+        new_value=location.coordinates,
+        notes=f"Re-forwarded capture GPS to location '{location.name}'",
+    )
+
+    db.commit()
+    return {
+        "success": True,
+        "location_id": location.id,
+        "location_name": location.name,
+        "coordinates": location.coordinates,
+    }
+
+
 # ── Helpers ───────────────────────────────────────────────────────────────────
 
 def _to_dict(pd: PendingDeployment, unit: Optional[RosterUnit] = None, detail: bool = False) -> dict: