Metadata-parser

- Several fixes to improve support of canonical and absolute urls
-- replaced REGEX parsing of urls with `urlparse` parsing and inspection; too many edge cases got in
-- refactored `MediaParser.absolute_url` , now proxies a call to new function `url_to_absolute_url`
-- refactored `MediaParser.get_discrete_url` , now cleaner and leaner.
-- refactored how some tests run, so there is cleaner output

- trying to fix some issues with distribution

- trying to parse unparsable pages was creating an error
-- `MetadataParser.init` now accepts `only_parse_file_extensions` -- list of the only file extensions to parse
-- `MetadataParser.init` now accepts `force_parse_invalid_content_type` -- forces to parse invalid content
-- `MetadataParser.fetch_url` will only parse "text/html" content by default

- trying to ensure we return a valid url in get_discrete_url()
- adding in some proper unit tests; migrating from the private demo's slowly ( the private demo's hit a lot of internal files and public urls ; wouldn't be proper to make these public )
- setting `self.url_actual = url` on __init__. this will get overridden on a `fetch`, but allows for a fallback on html docs passed through

- Dropped BS3 support
- test Python3 support ( support added by Paul Bonser [ https://github.com/pib ] )

- Pull Request - https://github.com/jvanasco/metadata_parser/pull/1
Credit to Paul Bonser [ https://github.com/pib ]