Support generic data urls rather than just CKAN resources as input

[At the moment](https://github.com/okfn/datapusher/blob/master/datapusher/jobs.py#L223) it seems datapusher gets passed resource id and then looks up data url.

That is metadata pushed to a job would be:

```
'metadata': {
'url': path-to-data-csv-or-xls
'resource_id': res_id, # resource to push to (a bonus would to have this auto-created
'set_url_type': data_dict.get('set_url_type', False)
}
```

Why do this?
- It seems to be to be simpler - don't have to spend time looking up the resource url from CKAN
- More general - can support data not yet in CKAN