Bug Fix for ASE 3.19.1 dependency

The ASE 3.19.1 MDMin object has nsteps attribute that stops the optimization if self.nsteps exceed the steps arg. Resetting the self.nsteps to 0 fix the problem

catlearn/optimize/mlmin.py

@@ -282,6 +282,7 @@ def run(self, fmax=0.05, steps=200, kernel='SQE', max_step=0.25,
 
  while ml_converged is False:
  ml_opt.run(fmax=fmax*0.90, steps=1)
+ ml_opt.nsteps = 0
  pos_ml = np.array(guess.positions).flatten()
  self.list_interesting_points.append(pos_ml)
  pos_ml = apply_mask([pos_ml],