dom/nodes/MutationObserver-characterData.html: ProcessingInstruction case still expects pre-whatwg/html#12118 bogus-comment parsing #63811

dom/nodes/MutationObserver-characterData.html has a subtest titled "characterData ProcessingInstruction: data mutations" whose node comes from the markup

<p id="n50"><?processing data?></p>

and whose first expected record is {type: "characterData", oldValue: "?processing data?"}.

That expectation dates from when the HTML parser turned <?...?> into a bogus comment, so n50.firstChild was a Comment with data ?processing data?. Despite its title, the subtest was testing a Comment.

Since whatwg/html#12118 ("Parse XML-style <?target data?> processing instructions", merged 2026-06-25), the parser creates a ProcessingInstruction with target processing and data data. Setting n50.data = "CHANGED" therefore records oldValue: "data", not "?processing data?".

Current results (wpt.fyi, aligned stable run at cc74d26):

• Chrome, which ships PI parsing: 22/23. It fails only this subtest, with the oldValue mismatch.
• Firefox and Safari, which still parse a bogus comment: 23/23.

Suggested fix: make the node independent of the parser, which also makes the subtest test what its title says. For example:

var n50 = document.createProcessingInstruction("processing", "data");

and expect oldValue: "data" for the first record. Alternatively, keep the markup and expect "data", which matches the current HTML spec.