From planning to policy: distilling $\texttt{Skill-RRT}$ for long-horizon prehensile and non-prehensile manipulation