Handle invalid blocks without modifying their markup

[nylen]

We need to detect blocks with invalid markup (e.g. added classnames, added or removed tags).

One way to do this (idea from @mtias): Detect "invalid" blocks by re-serializing blocks that we see in the editor. If it changes beyond a certain threshold (e.g. whitespace doesn't matter), then show the block as a core/html block and highlight how the markup would be changed.

Like #1735 it is important that the default "save" operation not alter the markup inside the block.

[afercia]

As an example: noticed during work on the Pullquote block, see #1987

When editing a post with a pullquote in the Classic editor, and then again in Gutenberg, the content gets lost. For example: inserted a pullquote with 2 paragraphs:

<blockquote class="wp-block-pullquote alignnone">
    <p>My quote</p>
    <p>with multiple par</p>
    <footer>my caption</footer>
</blockquote>

Saved, then opened the same post using the Classic editor, switched to text mode and the markup was changed to:

<!-- wp:core/pullquote -->
<blockquote class="wp-block-pullquote alignnone">My quote

with multiple par

<footer>my caption</footer></blockquote>
<!-- /wp:core/pullquote -->

This perfectly works in the Classic editor, but when editing again in Gutenberg, it doesn't of course and the content gets stripped out.