Support cache() & persist() of RDD

[AmbitionXiang]

I see there are cache.rs and cache_tracker.rs. But some functions, e.g., get_or_compute, are not used in the framework. Since persistence is also a core functionality provided by Spark, it's necessary to implement it.