Enable caching in MapInput for Spark

Because of the IOContext problem (HIVE-8920, HIVE-9084), RDD caching is currently disabled in MapInput. Prototyping shows that the problem can solved. Thus, we should formalize the prototype and enable the caching. A good query to test this is:

from (select * from dec union all select * from dec2) s
insert overwrite table dec3 select s.name, sum(s.value) group by s.name
insert overwrite table dec4 select s.name, s.value order by s.value;