Fixed annotations for borderline cases

Please refer to https://github.com/weblyzard/inscriptis/releases/tag/2.0rc1 for a list of all new features. This release candidate fixes the following issues in rc1:

• fixed annotations for some borderline cases
• improved documentation compared to 2.0rc2

Improved document model, parsing of borderline cases & HTML annotation support

1. HTML parsing:

   • new: new model for handling blocks and lines
   • chg: improved HTML parsing of tables, enumerations and margins; fixed borderline cases
   • chg: improved whitespace handling
   • add: cover more borderline cases with unit tests

2. Inscriptis core:

   • new: support for annotation rules and annotation output
   • new: annotation post-processors (html, xml, surface form)
   • new: type hints
   • chg: extended and improved documentation

3. Inscript command line client:

   • chg: apply --encoding to Web URLs as well

1.2

• tables: add support for vertical (valign, css: text-vertical-alginment) and horizontal (align) cell alignment (fixes: #33)
• improved handling of HTML attributes and styles
• code cleanup
• migrated build from travis to github actions

Improved margin handling & more liberal licensing

• ignore top margins at the beginning of a document.
• more liberal licensing:
  • the license change has been triggered by another project that created a Java port of inscriptis.
  • to facilitate the free sharing of code and ideas between our two projects, we have (i) obtained the permission of all contributors for a license change, and (ii) changed the inscriptis license to the "Apache License 2.0".

Improved testing and Python 3.9 support

• minor performance improvements and code optimizations
• added Python 3.9 test environment
• improved test coverage
• updated package metadata
• improved tox configuration

Improved HTML rendering, command line client and Web service

1. added support for rendering tags with the white-space: pre CSS attribute (e.g. <pre> which is often used for formatting code).
2. API change: A ParserConfig object replaces the parameters display_images, dedpulicate_captions, display_links and indentation in get_text() and for initializing the Inscriptis class.

      from lxml.html import fromstring
      from inscriptis.model.config import ParserConfig
      
      html_tree = fromstring(html)   
      # optional parser configuration fine tuning
      config = ParserConfig(display_links=True, display_anchors=True)
      parser = Inscriptis(html_tree, config)
      text = parser.get_text()

3. command line client:
   • added option for displaying anchor links
   • --encoding not sets the HTML and output encoding
   • new --version option
4. Web service
   • use the related CSS profile per default
   • added version call
5. Documentation fixes and improvements

Improved performance and code structure, documentation and unit testing

• improved performance and code structure.
• use metadata published in ./inscriptis/__init__.py for versioning and in setup.py.
• improved test coverage
• created sphinx API, usage and testing documentation which is published on https://inscriptis.readthedocs.org
• requires Python 3.5+ (dropped support for Python 2.7)

Correct inscript.py default indentation strategy.

Use the extended indentation strategy per default as outlined in the README.md.

Improved indentation and custom rendering styles

• improved indentation, if span and div tags are used
• support for custom rendering styles
• improved documentation
• use travis for auto CI
• requires Python 2.7+ or Python 3.5+ since lxml does not support Python 3 versions <3.5

Improved table rendering (nested tables and line breaks in tables)

• Correctly handle nested tables and line breaks (e.g. due to enumerations, list or paragraph breaks) in tables.
• Improved content stripping.

Please take a look at the Rendering document for an overview of how Inscriptis renders different tables.