Encapsulate all tabular read/write operations into tabulator

[roll]

Overview

Implementation of reading and writing tabular data is a error prone process. Especially writing in Python. It could be encapsulated to base library and re-used in many places (now almost the whole stack writes csv manually).

Tasks

• ensure all readings thru tabulator
• Write mode for topen tabulator-py#36
• update all higher level libraries and tools for writing using tabulator
• mimic this approach in other language implementations where tabulator is available

[pwalsh]

I don't see why we need this issue here, when we have this here frictionlessdata/tabulator-py#36

[roll]

I suppose the last usage of raw csv.write will be replaced here - frictionlessdata/datapackage-py#15. So closing the meta issue.