Add crude type parsing for jai tokenizer

[rainbowism]

It'll highlights the types in most situations. It isn't a very robust solution but after some testing it seems to work well enough.

[focus-editor]

Thanks for the PR! I tried to do it myself but decided against it because there are too many special cases (see screenshots)

I'm honestly not sure it's even worth it to try to highlight user types using tokens at this point. The plan was to leverage the compiler eventually to get a list of types asynchronously and put it in a hash map for highlighting (and other things)

[rainbowism]

That's fair. I have actually resolved those 2 scenarios but if the plan is for using the compiler rather than a bespoke tokenizer then that seems like a better plan.

[focus-editor]

There's more than 2 unfortunately. Also I checked and it slows down the parser by 5-10%, which is not great because we're parsing on every keystroke.

Here's some more (handling these cases would slow it down even further).

Would you agree to wait until we try it with the compiler?

[rainbowism]

That's fine with me. I can just use this personally for now in the meanwhile, or somebody can pull in my branch if they wish to use it themselves.

[focus-editor]

Great. Thank you for your understanding!