Add more copy-paste filtering

[thibaudcolas]

Follow-up to #50, #9.

The current implementation filters blocks, styles, entities by type. What formats it preserves (or how well it does that) depends a lot on the specific browser + OS + copy source being used. I've done lots of testing, and started collecting unfiltered ContentStates to be able to review and unit test this further.

First test run

Editor - Browser	Chrome Windows	Chrome macOS	Firefox Windows	Firefox macOS	Edge Windows	IE11 Windows	Safari macOS	Safari iOS	Chrome Android
Word 2016								iOS11
Word 2013		N/A		N/A			N/A	N/A	N/A
Word 2010	v62, Win8.1	N/A	v57, Win8.1	N/A		Win8.1	N/A	N/A	N/A
Apple Pages	N/A	v62, macOS10.13	N/A	v62, macOS10.13	N/A	N/A	v11, macOS10.13	iOS11	N/A
Google Docs	v62, Win8.1	v62, macOS10.13	v57, Win8.1	v62, macOS10.13	v16, Win10	Win8.1	v11, macOS10.13	iOS11
Word Online	v62, Win8.1	v62, macOS10.13	v57, Win8.1	v62, macOS10.13	v16, Win10	Unsupported	v11, macOS10.13	No paste	?
Dropbox Paper	v62, Win8.1	v62, macOS10.13	v57, Win8.1	v62, macOS10.13	v16, Win10	Unsupported	v11, macOS10.13	iOS11	?

Next steps

Based on a quick review,

• Drop line breaks
• Drop unneeded entity data
• Drop images based on criteria to be defined (data: URLs should be dropped, file: URLs should be dropped, potentially more).

Here are my detailed findings after reviewing the "Word 2010 on Chrome in Windows"

Special attention:

• Preserves line breaks
• Link entities have both href (source?) and url (added by Draft.js) attributes

Correctly retained

• Bold, italic, bold + italic, strikethrough, underline
• Unstyled, h1, h2, h3, h4, h5, h6
• Special chars (ellipsis, emojis, etc)
• "Link info" saved in title attribute

Not retained

• Superscript / subscript: unstyled
• Quote: unstyled
• Link target ("frame" in Word): no attribute for this
• Bulleted list: unstyled preceded with small middot bullet
• Nested bulleted list (1): unstyled preceded with medium empty circle bullet
• Nested bulleted list (2): unstyled preceded with special symbol
• Numbered list: unstyled preceded with (correct) count / letter (1., 2., a. for 1-level nesting)
• Star list: image with * in alt, and src set to local file

Incorrectly retained

• Outline: retained as bold
• Double strikethrough: retained as strikethrough
• Empty unstyled block inserted at the top
• Internal document link saved as link anchor in href attribute, and added by Draft.js at the end of url to point at the current page
  • eg. href: '#_Inline_styles'
• Comment reference text saved as [T.C1] (user initials)
• Comment saved as both link anchor in href attributes with special text on comment location, and end-of-document comment body with other special link anchor in href
  • eg. href: '#_msocom_1' for ref, href: '#_msoanchor_1' for body
• Image retained with height and width attributes (string repr of number)
• Image src is a file:/// protocol URL

[thibaudcolas]

See https://github.com/thibaudcolas/draftjs-paste for current progress. The test suite is going to be tens of MBs of JSON, and this isn't really related to Draftail, so I'm considering making a separate package.