Exporters

Fix excessive memory consumption in threaded mode

Fix HubStorageReader :
- Removed deprecated arg for collection scanner.

- Bugfix when decompressing multi-chunk compressed stream reader
- Experimental multi-threaded export

New features:
- Added retry reading streams 326 (partial fix for 322 )
- Added FlatsonTransform for easy flattening 324

New features:
- Introduced StreamBasedReader that adds support for custom decompressors and deserializers 316
- Added support for uncompressed input via NoDecompressor 316
- Added support for CSV formatted input through CSVDeserializer 316
- FSReader now ignores directories whose name start with a dot 314
- Added DupeFilter, a filter that ignores subsequent duplicates based on a key field 321
- Allow using integer as project id 309

Bug fixes:
- Support long values in int fields 320

- Bugfix that prevents the bypass code to swallow exceptions
- Updates setup.py metadata
- Adds license info
- dicttoxml and bz2file are now lazy requirements