Use parallel=True in rolling.apply

[mroeschke]

Gains an additional speedup from #29

In [1]: s = pd.Series(range(10000))

In [2]: r = s.rolling(10)

In [3]: f = lambda x: np.sum(x) + 5

In [4]: %timeit r.apply(f, raw=False)
656 µs ± 112 µs per loop (mean ± std. dev. of 7 runs, 1 loop each)