Sockit

Corresponds with [sockit](https://github.com/ripl-org/sockit/) release v0.3.0

A minor release to fix the broken PyPI description and logo image in 0.2.0.

- Significant refactoring to include additional functionality for parsing job postings and resumes (`sockit parse`) and comparing job postings, resumes and SOC codes (`sockit compare`).
- Original functionality for probabilistic matching of job titles to SOC codes has been moved under `sockit title`.
- Data has been expanded from 2019 NLx Research Hub job postings to 2019 and 2021, and from a single wordtrie of job titles to many additional files under the `data` directory (skills, topic-modeling matrices, acronyms, abbreviations, resume headings, schools, degrees, and fields of study).

- Adds scoring mechanism to weight matches based on title words that match to a list of job title nouns and to nodes in the job title trie

- Refactored result sorting code into new `sort()` method.
- Fixed normalization error when search results are empty.
- Rebuilt wordtrie with stopword filtering (mainly removes place names).
- Support for JSON input files.
- Added test cases and a script for comparing test results across different implementations.

Initial release.