Parallelize `prescrump` and `prescraamp`

[seanlaw]

Currently, stumpy.prescrump is single threaded. However, with some clever bookkeeping and engineering, it may be possible to parallelize this section of the code with numba:

for i in np.random.permutation(range(0, l, s)).astype(np.int64):
        QT = core.sliding_dot_product(T_A[i : i + m], T_B)
        _prescrump(
            T_A,
            T_B,
            m,
            M_T,
            Σ_T,
            μ_Q,
            σ_Q,
            QT,
            i,
            s,
            squared_distance_profile,
            P_squared,
            I,
            excl_zone,
        )

With the new core._sliding_dot_product, which is njit enabled, should be able to parallelize this.