fix: stop double encoding query params

[SethFalco]

While working on #458, I noticed something strange with the URLs.

A line break was encoding to %250A, and a space would encode to %2520, which doesn't seem right. Turns out special characters were being double encoded/decoded. This is mostly a waste of resources, but also increases the likelihood of a user running into #437 because it inflates the URL.

It's also why all special characters had an encoding starting with %25. Because they were encoded already, so % is the only possible special character, and then the % got URI encoded to %25.

URLSearchParams already does URI encoding and decoding under the hood in #set and #get. There's no need for us to do it!

Here is a demonstration of the issue:

const char = "<";
const encode = encodeURI(char);

const bad = encodeURI(encode); 
// Result: %253C

const params = new URLSearchParams();
params.set("q", encode); 
const bad2 = params.toString();
// Result: q=%253C

On libretranslate.com, if you insert < in the translation textarea, it currently becomes %253C which is not the expected URI encoded string. It should be %3C!

const query = new URLSearchParams(); 
query.set("q", "<");
const good = query.toString();
// Result: q=%3C

const good2 = query.get("q");
// Result: <

[pierotofy]

Yep, you're right, looks like we're unnecessarily encoding/decoding here. Good catch!