ekf2: remove sparse vector optimization

The sparse vector template requires to know which states are non-zero in
the observation jacobian. This complicates the modularity of the code
when the state vector or the derivation is changed.
The computation cost difference is almost negligible for this size.

src/modules/ekf2/EKF/ekf.h

@@ -63,9 +63,6 @@ public:
 	typedef matrix::Vector<float, State::size> VectorState;
 	typedef matrix::SquareMatrix<float, State::size> SquareMatrixState;
 	typedef matrix::SquareMatrix<float, 2> Matrix2f;
-	template<int ... Idxs>
-
-	using SparseVectorState = matrix::SparseVectorf<State::size, Idxs...>;
 
 	Ekf()
 	{