Remove regex use from gleam/uri

[lpil]

Convert it to manually parse the URI instead.

[giacomocavalieri]

I started looking into this and noticed a couple of things with the js impl: it's using a variation of this regex ^(([^:/?#]+):)?(//([^/?#]*))?([^?#]*)(\?([^#]*))?(#(.*))? provided in the Uri rfc. However this will behave differently from the implementation for the Erlang target (done by uri_string:parse); for example :wibble is rejected on the Erlang target but is parsed as a uri with path ":wibble" on the js one

I'm pretty sure it's the Erlang implementation that is correct here since that regex is only meant to "break down a well-formed URI reference into its components". In any way it would be best if both targets behaved the same here.

• If the plan is to get rid of the regex dependency to v1 stdlib as soon as possible I can go ahead and implement a parser equivalent to that regex, it would take very little time but the target differences would still be there!
• Or since we're making this change we could try and copy Erlang's implementation but that would be a big item of work (the erlang implementation is thousands of lines of code from what I can tell and understanding it and porting it to Gleam could require a long time)

What do you think is best here?

[lpil]

Removing the regex and making an issue for making it consistent between both targets sounds find to me.