Compute license score on 5K

[pombredanne]

For this larger run, we should still keep the JSON scans... but only that and not the CSVs.
Also as soon as the data is satisfying enough we can push a ScanCode update to CD and see how we can trigger a rescan on the 5k popular packages with the new ScanCode options.

[pombredanne]

Here is an update to the License score status:

• we completed a scoring run on 5K pacakges... and will post this soon online
  • only about 150 that have a score of 80 or up
  • this is the sad state on license clarity and also the fact that the score gives penalties quickly as most elements are binary elements
  • if we focus the license text element only on the key files, this will be worse

next steps:

• discuss and validate:
  • with other CD folks: call this or next week?
  • on the license-discuss OSI list and FSFE LN list?

integrate in the CD codebase:

• update definition schema to support
  • a "contains license text" data element
  • each scoring elements as stored data
• compute the score proper

communicate on the score:

• article to be written, either as me or as CD project? TBD
• presentation at Paris Open source summit (2018-12)
• FOSDEM talk proposal