Improve diff readability for eg "Bob," -> "Bob"

[laughinghan]

If a hunk consists of only deleting or only adding at the beginning
or end of a word, then combine them into one hunk.

Examples:

• Changes from simplediff:

  - Alice, Bob and Charlie
  + Alice, Bob, and Charlie

  simplediff: Alice, BobBob, Charlie
  this diff: Alice, Bob, Charlie

  - Alice Bob Charlie's Angels And David
  + Alice Bob Charlie David

  simplediff: Alice Bob Charlie's Angels AndCharlie David
  this diff: Alice Bob Charlie's Angels And David

• Same as simplediff:

  • hunks you wouldn't want simplified:

    - Alice Bob Charlie
    + Alice Robert Charlie

    diff: Alice Bob Robert Charlie

  • if the change isn't only at the beginning or end:

    - Alice Bob Charlie
    + Alice Blob Charlie

    diff: Alice Bob Blob Charlie

    - Alice Bobby Charlie
    + Alice bb Charlie

    diff: Alice Bobby bb Charlie

    - Alice Zeneca Charlie
    + Alice AstraZeneca Charlie's

    diff: Alice Zeneca Charlie AstraZeneca Charlie's

[j-e-d]

Thanks for the PR @laughinghan, let me find some time to test it locally and I will merge it if it looks good.

[j-e-d]

@laughinghan the problem I have with your proposed patch is with cases like the following:

old = "Provide them Clues"
new = "Provide the Clues"
    
assert html_diff_original(old, new) == 'Provide <del>them</del> <ins>the</ins> Clues'
assert html_diff_proposed(old, new) == 'Provide <del>them</del> <ins>the</ins> Clues'

The assert with the current html_diff is what I would expect, the word was changed from "them" to "the" and I want the diff to show the whole word, what I get when I run that test with the proposed change is:

E       AssertionError: assert 'Provide the<...m</del> Clues' == 'Provide <del...e</ins> Clues'
E         - Provide <del>them</del> <ins>the</ins> Clues
E         + Provide the<del>m</del> Clues

I would be OK and like that the algorithm didn't consider punctuation marks as part of a word, but I don't want words to be cut into parts when they are changed.