Parser: Fix JS/PHP inconsistencies with empty attributes

[dmsnell]

This patch updates a couple of inconsistent behaviors when dealing with
empty attributes and with the way PHP and JS treat empty objects differently.

PHP empty attributes

When returning blocks without attributes the spec PHP parser has been
sending [] instead of {} due to complications of how PHP serializes
empty arrays to JSON.

In this patch we're adding a new test to verify the behavior and then
fixing the spec parser so that the output from the PHP version matches
the output from the JS version identically.

Bug with empty attributes in comments

The default parser introduced a bug where <!-- wp:anything {} /--> would
fail to parse because it expected some content inside the curly brackets for
the JSOn attributes.

This is fixed in this patch by changing the +? quantifier in the RegExp
tokenizer with a *? to allow for empty attributes.

A test suite has been added to verify that when we parse an array of different
block formulations that they produce what we expect. A bug in a test was the
reason I didn't originally find this bug.

[mcsf]

I was about to approve this, but in the course of testing I found an issue that is also present in master:

Open Gutenberg's Code Editor, paste the following:

<!-- wp:foo {"id":824} -->
.
<!-- /wp:foo -->

<!-- wp:foo {} -->
.
<!-- /wp:foo -->

<!-- wp:foo -->
.
<!-- /wp:foo -->

I expected this to be preserved verbatim or at the very least to have the middle block "self-correct" by deleting its empty attributes object. Instead, upon blurring the code editor, I get this:

<!-- wp:foo {"id":824} -->
.
<!-- /wp:foo -->

<p><!-- wp:foo {} --><br>
.<br>
<!-- /wp:foo --></p>
<p><!-- wp:foo --><br>
.<br>
<!-- /wp:foo --></p>

Comparing the two spec and default parsers:

---

I assume this to be independent of the current PR, so it's up to you whether we should address it separately. In any case, 👍.

[mcsf]

^ to be clear, I'm not concerned about the parser discrepancies when it comes to collecting empty freeform nodes — just the missing core/foo blocks. :)

[dmsnell]

Thanks for pointing out that bug @mcsf - I saw that a day or two ago and forgot to submit a new PR to fix it - #11690

I believe this is due to demanding that attributes aren't empty, but there's no reason they can't be.

[dmsnell]

Squashed and rebased; former HEAD was 052c689

[dmsnell]

@mcsf I folded the {} fix into this PR since this is where all the tests are introduced. should be fixed for you now - thanks again for pointing that out.

[gwwar]

With d1acdf6 @mcsf are we good to 🚢 ?

[pento]

👍🏻 Looks good!