Store schemas in storage #27
Description
Requires frictionlessdata/tableschema-py#59
On previous iterations I've run into some problems using 100% pure reflection system. And now it's obvious for me that for first-class JSON Table Schema features support we need to store schemas in the storage (in the meta table _schemas for example). It's not against current design and could be implemented relatively fast.
For tables not having _schemas current reflection system is still be working with subset of features supported. So user still will be able to describe and read just a SQL table (the same as now).
Really related - frictionlessdata/datapackage-py#69