performance-only fixes (np.clip, np.array growing, )

[michaelfeil]

as discussed in #101:

• np.clip is slower than np.core.umath.clip (saves ~10%) simulation time
• lists instead of np.array growing and smaller fixes (saves ~10% simulation time)

Time before for 10 runs for pusing task (200 steps,3 resets) : ~13.0s +- 200msafter 10.8s +- 150ms

All changed functions have been like: np.testing.assert_array_equal(prior_implementation, faster_implementation) while running the pytests