Major Description
Currently the Lucene index loads all paths in memory.
This could result in an out-of-memory if there are many results. Also, it slows down performance if we only care about some of the results.
Instead, the data should be loaded on demand if possible.
As we anyway refactor the full-text engine, it doesn't make sense to solve this problem right now (with the existing Lucene index implementation), but wait for OAK-902 to be completed.