Make Delta.diff() match the quill-delta.js output, returning a leaner diff in some cases

[roughike]

What & why

Right now, in certain cases, Delta.diff() returns more stuff than it needs to.

For example, in the following case:

final a = Delta()
  ..insert('Test\n', {'index': 0})
  ..insert({'image': 'https://example.com/image.png'}, {'index': 1})
  ..insert('aaa\n', {'index': 2});

final b = Delta()
  ..insert('Test\n', {'index': 0})
  ..insert('\n', {'index': 0.5})
  ..insert({'image': 'https://example.com/image.png'}, {'index': 1})
  ..insert('aaa\n', {'index': 2});

print(a.diff(b));

the expected output is:

retain⟨ 5 ⟩
insert⟨ ⏎ ⟩ + {index: 0.5}

but the actual output is:

retain⟨ 5 ⟩
insert⟨ ⏎ ⟩ + {index: 0.5}
insert⟨ {image: https://example.com/image.png} ⟩ + {index: 1}
delete⟨ 1 ⟩

The last insert and delete are redundant. For some reason, our OT implementation will break with the current behavior (the document will contain two images instead of one), but this PR fixes that. The actual problem might be somewhere else, but I think it would be still ideal if this Dart quill-delta library worked the same way as the JS equivalent.

To compare with the output of the JS quill-delta library, I've set up a Replit you can verify this with here.

[codecov]

Codecov Report

Merging #27 (2769cd6) into master (4846633) will increase coverage by 0.32%.
The diff coverage is 100.00%.

❗ Current head 2769cd6 differs from pull request most recent head c267ddf. Consider uploading reports for the commit c267ddf to get more accurate results

@@            Coverage Diff             @@
##           master      #27      +/-   ##
==========================================
+ Coverage   93.56%   93.89%   +0.32%     
==========================================
  Files           1        1              
  Lines         342      344       +2     
==========================================
+ Hits          320      323       +3     
+ Misses         22       21       -1     

Impacted Files	Coverage Δ
lib/quill_delta.dart	93.89% <100.00%> (+0.32%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 4846633...c267ddf. Read the comment docs.

[roughike]

@pulyaevskiy the CI is not passing because of formatting. I didn't want to create a bigger diff than it needs to be, but let me know how I should proceed :-)