JSON key `"id"` breaks the tokenizer

The internal tokenizer uses the key "id" as a way to keep track of rules within a grammar
However, the parser does not strip "id" keys from the grammar before parsing
if a invalid "id" key is hit during tokenizing, it will lead to unpredictable tokenizing and the grammar partially failing

Luckily it seems no one has placed one in the wrong spot (having one at root is fine)
tho I assume that's because they would have noticed their grammar no longer working when it is present
https://github.com/search?q=path%3A*.tmLanguage.json+%22%5C%22id%5C%22%3A%22&type=code

{
	"scopeName": "source.lang",
	"patterns": [
		{
			"begin": "\"",
			"end": "\"",
			"name": "string",
			"patterns": [ { "include": "#backslash" } ]
		}
	],
	"repository": {
		"backslash": {
			"match": "\\\\.",
			"name": "constant.character.escape",
			"id": "JSON key \"id\" breaks this rule and its parent"
		}
	}
}

No source file or test is named. Start by tracing the grammar parser and internal tokenizer handling of "id" keys, using the JSON grammar example as a reproduction. Add a regression case showing that an "id" inside a rule does not corrupt tokenization or cause the grammar to partially fail.