[ocr-parser][l]: parser for ocr

[osahon-okungbowa]

ocr-parser complete. Completed the following:

• I had to update the dump() for the Dataset model it allow for the creation of multiple unique json files from the same source_url. This is necessary to cater for pages that produce multiple datasets per page.

• I also added kwargs to the Dataset & Resource model init() to allow me instantiate objects with attributes not yet defined. I didn't do anything with the kwargs.

• I added a 'parsers' package to the 'ocr' package and a 'parser' module to the 'base' package

[nightsh]

I have removed the old usage of dataset model, since as of yesterday we started using the scrapy pipeline. So no more dump() calls.

Here's a summary of my changes:

• no more dataset_list in the parsers - replaced by yield dataset whenever a dataset model is completed
• no mo home made slugs, imported slugify to make slugs
• name of the dataset is a slug of its title, not of its source_url
• both Dataset and Resource objects are now subclasses of python dict (Scrapy.Item, to be more specific) - this is mostly for compliance with the Scrapy way of doing things, but also helps us a bit around the code. This time is made more sense to have it than on my first try.
• since the models extend Item now, no more need for **kwargs, it's built in :-)
• changed the file name from using timestamp to using an additional hash of name - so it will overwrite old results on subsequent runs, instead of creating new results

I am going to run the code from this branch now (i.e. not merge yet), so you can have a look at the diff and maybe have suggestions.