TermVectorAccessor return terms that do not match with current document [LUCENE-4720]

[asfimport]

For each term, TermVectorAccessor looks in the indexReader and calls termPositions.skipTo(documentNumber) but this methods returns the first document with id greater or equal to documentNumber.
As a result you get some extra terms that do not really match with documentNumber.

---

Migrated from LUCENE-4720 by Francois-Xavier Bonnet
Attachments: SOLR-4360.txt

[asfimport]

Francois-Xavier Bonnet (migrated from JIRA)

Here is a patch