support for non-sequential TSV data (for Fintan)

[chiarcos]

When processing CoNLL, we need to keep track of sequence (nif:nextWord, nif:nextSentence) and structure (nif:Word, nif:Sentence, conll:HEAD).
When processing one-word-per-line dictionary formats (say, Unimorph, TIAD-TSV or OMW-TSV), sequence and groups don't matter anymore.

Request:
Create TSVStreamExtractor as a copy of CoNLLStreamExtractor that
(a) does not produce NIF properties and classes, that
(b) replaces all conll: properties with properties in a Fintan namespace
(c) processes line by line or a fixed (conigurable) number of lines
(d) separates the RDF output from each (sequence of input) line(s) with one empty line

Define CoNLLStreamExtractor as a subclass of the TSVStreamExtractor.
Overrides Fintan behavior (see above) with current behavior.

Create FintanUpdater as a copy of CoNLLRDFUpdater, define CoNLLRDFUpdater as a subclass of FintanUpdater

• should support SPARQL SELECT statements to export TSV directly (without CoNLLRDFFormatter)
• keep CoNLLRDFUpdater class for backward compatiblity, no difference in functionality