The non-tree sitter version is almost completely copied so nothing to say there.

There are some inconsistencies/missing operators in the tree-sitter version, for example:
- `this`
- `new.target`
- `import.meta`
- The keyword.operator order looks more wrong the more I look at it but that's for another day because of the possible scope regularity change in the future: atom/language-javascript#690
- btw atom/language-javascript#691 looks good to me I should probably do that

Even though tree-sitter will have some above changes I added the new operators (`** **= ?? &&= ||= ??= void`) anyway.
Well `void` is not new it's just missing.
See https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators for a list of operators. Of course while the comma is technically an operator, hopefully it's used more to separate elements in arrays or properties in objects.

Also `delete` is an operator not control.

?. is illegal in new

update console methods
timeStamp has the same support as profile and both are nonstandard so they're both included

update global classes
Wait a minute, Generator and GeneratorFunction are _not_ global...
so actually just any class specified in ECMA262
non global additions:
- AsyncFunction
- AsyncGeneratorFunction
- TypedArray
luckily MDN's sidebar of
https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/globalThis
shows all of these
Wait... JSON? That's not a class
Math has the scope support.class.math.js for some reason so I guess add "json" likewise for json

Never knew about WeakRef and FinalizationRegistry
Wonder why it's so unknown; this is a common idea in Rust at least....

So yeah, JSON is separate just like Math
Lots of copying and pasting

rest vs spread depends on context
I don't know why that test failed

given that the entire Relay.QL was entity.name.function, it's not surprising here
Therefore this pr just adds "meta.delimiter.<>" to the delimiter in Relay.QL or Relay?.QL