Migrate to native `URL`

[pi0]

Historically ufo was started as a polyfill for URL when there were no enough browser coverage and also due to behavior variations. Now that URL is an standard de-facto, we can migrate to it for all (for parse/stringify) or at least for resolve/normalize Utils.

One of the biggest breaking change side-effects of this is that we will encode more characters even then they not not necessary and parsable. The initial behavior was taken from vue-router that preferred to keep URL readability and encode in must-have situations only. (one example: #164)

Second thing to consider is the performance part which probably worth a decent benchmark against URL implementation of major JS runtimes (Browsers, Node, Bun, Deno, Cloudflare workers) comparing to current regex based parseURL.

--

Update: I am thinking if we could ship it even faster with ufo/url or ufo/web to allow progressive opt-in and later in v2, have backward compact with ufo/compat. Only have to figure out a way to make bundling work to switch impl of encoding utils.

[pi0]

I get it from the Mulesoft - https://help.mulesoft.com/s/article/java-lang-IllegalArgumentException-Malformed-URI-while-migrating-to-Mule-4
Mulesoft still supports RFC 1738 where "{", "}", "|", "", "^", "~", "[", "]", and "`" are unsafe characters.

Originally posted by @enkot in #165 (comment)

[johnpyp]

+1 on ufo/url, that'd be great.

[Exotelis]

@pi0 Do you have any idea when this might happen?

I’m asking because the ` character isn’t being encoded correctly either, which is causing issues in ofetch (the default fetch API in Nitro). I would have expected the library to handle this out of the box, otherwise the extra package doesn’t add much value if I still need to encode user input manually.

Would you recommend that I open a separate PR for this?