FWIW, this was the code used to compare htmlunit vs Chrome behaviour of URLSeachParams's splitting. This PR fixes all the differences and I've rolled the important bits into the URLSeachParamsTest.appendSpecialChars() test.

<!DOCTYPE html>
<html>
<head>
<script>
var param = new URLSearchParams();
param.append("?x=1&", "foo");
param.append("bar", "http://foo.com/?x=1&y=2&z=3")
param.append("foobar", "baz")
// chrome: %3Fx%3D1%26=foo&bar=http%3A%2F%2Ffoo.com%2F%3Fx%3D1%26y%3D2%26z%3D3&foobar=baz
// htmlunit: x=1&=foo&bar=http%3A%2F%2Ffoo.com%2F%3Fx%3D1&y=2&z=3&foobar=baz
console.log(param.toString());
// chrome: true
// htmlunit: false
console.log(param.has("?x=1&"));
// chrome: http://foo.com/?x=1&y=2&z=3
// htmlunit: http://foo.com/?x=1
console.log(param.get("bar"));
param.set("foobar", "foo?bar&");
// chrome: foo?bar&
// htmlunit: foo?bar
console.log(param.get("foobar"));
param.delete("?x=1&");
// chrome: bar=http%3A%2F%2Ffoo.com%2F%3Fx%3D1%26y%3D2%26z%3D3&foobar=foo%3Fbar%26
// htmlunit: x=1&=foo&bar=http%3A%2F%2Ffoo.com%2F%3Fx%3D1&y=2&z=3&foobar=foo%3Fbar
console.log(param.toString());
</script>
</head>
<body>
</body>
</html>