[BUG] Fix reset() method for v2

[gunnxx]

reset() method does not reset the self._prev_obs. It creates inconsistent reset state which is supposed to be deterministic for the same tasks in the same environment (by task, I mean parametric variation).

import metaworld

benchmark = metaworld.MT1("reach-v2", 0)
env = benchmark.train_classes["reach-v2"]()
env.set_task(benchmark.train_tasks[0])

reset_1 = env.reset()
reset_2 = env.reset()
_ = env.step(env.action_space.sample())
reset_3 = env.reset()
reset_4 = env.reset()

print(reset_1 == reset_2)
"""
[ True  True  True  True  True  True  True  True  True  True  True  True
  True  True  True  True  True  True  True  True  True  True  True  True
  True  True  True  True  True  True  True  True  True  True  True  True
  True  True  True]
"""

print(reset_1 == reset_3)
"""
[ True  True  True  True  True  True  True  True  True  True  True  True
  True  True  True  True  True  True False False False  True False False
 False False False False False  True  True  True  True  True  True  True
  True  True  True]
"""

print(reset_1 == reset_4)
"""
[ True  True  True  True  True  True  True  True  True  True  True  True
  True  True  True  True  True  True  True  True  True  True  True  True
  True  True  True  True  True  True  True  True  True  True  True  True
  True  True  True]
"""

The following changes will return all True for reset_1 == reset_3.

[reginald-mclean]

@gunnxx Thank you for catching this! As Meta-World ownership has been transferred from the RLWorkGroup to Farama Foundation, I am closing this issue but it will be fixed in a future version of Meta-World. If there are any questions or requests for features please join our Discord