Implement faster distance algorithm (#125)

Implement `faster_current` distance algorithm

vrplib/parse/parse_distances.py

@@ -87,10 +87,13 @@ def pairwise_euclidean(coords: np.ndarray) -> np.ndarray:
  An n-by-n Euclidean distances matrix.
 
  """
- diff = coords[:, np.newaxis, :] - coords
- square_diff = diff**2
- square_dist = np.sum(square_diff, axis=-1)
- return np.sqrt(square_dist)
+ coords = np.atleast_2d(coords)
+
+ sq_sum = (coords**2).sum(axis=1)
+ sq_dist = np.add.outer(sq_sum, sq_sum) - 2 * (coords @ coords.T)
+ np.fill_diagonal(sq_dist, 0) # avoids minor numerical issues
+
+ return np.sqrt(sq_dist)
 
 
 def from_lower_row(triangular: np.ndarray) -> np.ndarray: