ekf2: predict covariance before predicting state (#20715)

* To predict the covariance, the jacobian from the previous to the current state is required.

src/modules/ekf2/EKF/ekf.cpp

@@ -103,8 +103,8 @@ bool Ekf::update()
 	// Only run the filter if IMU data in the buffer has been updated
 	if (_imu_updated) {
 		// perform state and covariance prediction for the main filter
-		predictState();
 		predictCovariance();
+		predictState();
 
 		// control fusion of observation data
 		controlFusionModes();