Migrate logic into paperetl and paperai projects

• Modify embeddings index logic to stream numpy arrays to temporary storage instead of storing fully in memory
• Reorganized section queries

• Moved ETL process from this project to https://github.com/neuml/paperetl
• Modified license from MIT to Apache 2.0

• New feature to add must query clause to queries. Similarity results must contain a token with a plus at the front of it within a query.
• Subquery system enhancements, bug fixes and improvements
• Updated task queries/subqueries

Fix report column names, removed debug prints, removed unused report methods.

Fix issue with report sorting

Major rewrite of the reporting framework to handle subqueries and more complex queries.

Reporting

• Report queries now backed by YAML files to allow more control over query structure. Reports can include standard columns as well as user generated columns.

• User generated columns can either be a constant value or subquery driven. Subqueries are backed by a transformers question-answering model that allows extraction of specific values within search results.

Database schema

• Removed stats table as logic is now available via report subqueries

Fixes and improvements

• Forked pymagnitude repository to simplify build process and remove dependencies/functionality not needed by this project. Removed related code no longer needed.
• Consolidated section query and filtering logic
• Refresh with data through 2020-06-02

Refresh with data through 2020-06-01

Refresh with data through 2020-05-27

Refresh with data through 2020-05-26