Ddf-utils

- use dask and categorical types to improve memory usage (90)
- sorting column/row order before serving result in Chef
- add `breakpoint` option to chef procedures to stop and debug chef
- add data/metadata downloading functions to various data sources
- minor changes and improvments

- new support Hy mode
- bug fixes and improvements

- add support for reading mutiple indicators in one DDF datapoints file (76)
- add support for reading repo from github / local path (79)
- add dry_run option in Ingredient to speed up loading speed
- get_datapoints_df() for Dataset object always return a DataFrame
- to_graph() for Chef now includes procedure names
- bug fixes and improvements

- support mongo-like queries in ingredient definition and filter procedures (72)
- new procedure: filter (72)
- inline ingredients in ingredients section and procedures (36)
- external csv file as ingredient (36)
- colourful terminal output for chef
- bug fix and minor improvements

- totally rewrite the ddf reader, make separated models for dataset and datapackage in
`ddf_utils.models`.
- add Chef class, which can create or load recipes interactively. Now all recipe related
tasks should run under a Chef instance.
- all procedures takes a chef instance as first arguments now. Also we support writing
custom procedures now.
- removed `indentity` and `copy` procedure, updated `filter_row` and `groupby` procedure
- added new `flatten` procedure
- minor improvements

- bug fix: 67