An encoding test does not appear to match the RFC?

The [first `ID_Encoding` test](https://github.com/google/robotstxt/blob/455b1583103d13ad88fe526bc058d6b9f3309215/robots_test.cc#L399) caught me by surprise, since it does not _appear_ to match the RFC:

```c++
// /foo/bar?baz=http://foo.bar stays unencoded.
{
const absl::string_view robotstxt =
"User-agent: FooBot\n"
"Disallow: /\n"
"Allow: /foo/bar?qux=taz&baz=http://foo.bar?tar&par\n";
EXPECT_TRUE(IsUserAgentAllowed(
robotstxt, "FooBot",
"http://foo.bar/foo/bar?qux=taz&baz=http://foo.bar?tar&par"));
}
```

However, [section 2.2.2 of the REP RFC](https://datatracker.ietf.org/doc/html/rfc9309#section-2.2.2) seems to indicate that `/foo/bar?baz=http://foo.bar` should be encoded as `/foo/bar?baz=http%3A%2F%2Ffoo.bar`.

I can't decide if I'm mis-reading the RFC or if the test intentionally deviates from the RFC in this case.

Thanks!