It provides hardcoded ES mapping, considering data inspection on 85 million documents would take long time
It enables reading processed, clean SemMedDB predication data frame when the cache file is available, considering the whole data cleaning process would take ~20min
This PR mainly fixes Issue https://github.com/biothings/semmeddb/issues/10.
In addition: