allow zig js host functions to return JSError

[nektro]

progress towards zig JSValue stability and safety in #159
more followups to come

[robobun]

❌ @nektro, your commit 891daa2 has 3 failures in #6080:

test/js/bun/http/serve.test.ts - 1 failing on 🍎 x64

test/js/web/fetch/fetch.tls.test.ts - 1 failing on 🍎 aarch64

test/regression/issue/ctrl-c.test.ts - 1 failing on 🍎 aarch64

[paperdave]

should we actually be annotating every function with bun.JSError!? i think my solution to this in dave/more-bake is better, since it doesnt include the catch control flow on functions that don't return an error

[nektro]

nearly every function annotated will be taking advantage of it eventually and unless there's a noticeable perf effect then the consistency makes the function's purpose very clear at a glance

a big motivator for this project is to make the code interacting with these values as predictable as possible to interact with so that it becomes hard or impossible to introduce certain kinds of bugs

[paperdave]

if we want all of these functions to have the explicit annotation, it should be bun.OOM || bun.JSError, which i have in my branch defined as bun.JSOOM (can be named something else). the wrapper checks and handles both of these errors, if specified.

this is why i asked to wait a day for my branch to finish, as i'm nearly close to merging it. was this blocking something else?

[nektro]

in a couple more stages i think it would make more sense for OutOfMemory to be a part of bun.JSError, without a union
pr is only blocking itself

edit: in a few more after that global.throw* will return JSError and we can use global.throwOutOfMemory()

[dylan-conway]

I think we should consider keeping this here and rename error.JSError to error.ThrownException or something similar

[nektro]

yeah there's many followups ahead of this