Support async db connector, convert project to use python 3.6, other changes

Version 0.8

Added

• support for asynchronous querying (SPARQL database),
• custom implementation of dict for tokens (TokenDict) as a workaround for an issue that tokens are not hashable in corpus2-python3.6,

Changed

• converting project from 2.7 to 3.6 version of Python,
• AllegroGraphDao now uses new sparql_connectors package, which provides asynchronous connector and neat conversion from db format,
• maintain compatibility with Neo4j database (updates in neo4j dao class, added default implementations for batch variants of methods),
• refactor of methods in elinker module (required because of changes):
  • introducing batch querying as it is more beneficial in case of async connector,
  • other changes,
• update list of known LOD resources,
• implementation of method returning unannotated tokens,

Fixed

• Fixed crosswiki resource file: decoding (%-escaped) entity identifiers and using English identifiers instead of Polish,
• Decreasing execution time by fixung caching and adding extra info to get_annotations function call to avoid expensive operations,
• Change SPARQL queries: improve (proper to current structure of AG database) and fix (as current queries were incorrect),
• better resolving of lemmas for crosswiki links:
  • wsd: if there is no crosswiki entity for lemma, then preferred lexeme is checked,
  • mwe and ne: using proper base lemmas of annotations (as merging of base lemmas from tokens is not reliable),
• Remove filtering statement for LOD resources,