Avoid loading EFO ontology for each PMID in Data Release

[ala-ebi]

In the DR, each PMID job loads the ontology in memory before starting indexing, which takes time and sometimes takes longer than the indexing itself, we can eliminate this in different ways, one way is bundling multiple pmids in in the same job, the indexer already supports that using -p pmid1,pmid2,pmid3...
ideally we need some intelligence to figure out which pmids to bundle together in the same job, we dont want multiple big pmids in the same job.