Fix content attribute selector

[yscik]

Follow-up to #163

Changes proposed in this Pull Request

• Change back the selector for content attribute, since the extra <code> wrapper from Fix escaping issues with HTML entities #160 was not needed after all.
  • This also caused an issue of the block being re-created from the deprecated helper on every open
• This also removes the need for the deprecated helper — this would make sure block output from older versions are compatible, but the block output is not going to change in 3.5.6

Testing instructions

• Create a code block in master
• Switch to this branch
• Edit the post and ensure the block works without any warnings
• Change/add new code blocks, save post.
• Reload the block editor and check that block still works without warning.

[jom]

Something strange happens with this when coming from 3.5.5.

I created a block with the test content from #152. I then switched to this branch, edited the page, saved it, and escaping seems to have been lost.

This didn't happen in the page I created on master. Is this expected with how content escaping is changing?

[yscik]

The fix in #160 is implemented when saving the block, so it won't work with existing content from 3.5.5 — it will still break the content when loading the non-escaped output of a 3.5.5 block. When the post content is parsed, both ¢ and ¢ will be read as ¢.

I tried an approach of escaping the block content on the REST API before it hits the editor, which could also fix this issue for existing content. Even though the block would pick up the correct content, block validation then fails because the post content (escaped) is not the same as the save output (not escaped) 😫 . This would likely need setting up a block version attribute and conditionally escaping, but it's a pretty fragile solution already.

Considering that these string in existing content are likely already broken if it was edited even once, it might not be worth it to complicate this further. cc @alexsanford any opinions about this?

[jom]

Needs a rebase to work with the refactor, but I'm happy with this approach 😃