fix(grpc): Fix handling of non-UTF-8 strings in target URI path

[arjan-bal]

In Rust, standard strings must hold valid UTF-8 data. While there are types to represent non-UTF-8 strings in the standard library, they are currently unstable.

This PR ensures that non-UTF-8 strings are correctly parsed from the target URI path and propagated through the channel.

Key Changes:

• Updates ByteStr semantics: The UTF-8 invariant on the ByteStr type in the gRPC crate has been dropped, allowing it to store arbitrary bytes. I also implemented common traits to support string-like operations and cross-type comparisons. This enables resolver Address types to hold arbitrary byte sequences.
• Modified Target::path(): This function now percent-decodes the path and returns a ByteStr instead of an &str.
• Updated Unix socket tests: Because Unix file paths can contain non-UTF-8 characters, the tests have been updated to verify that these byte sequences propagate correctly through the channel.

Notes on Authority and UTF-8:

The Authority struct and the channel option's override_authority function continue to use standard strings, as they map to the HTTP/2 :authority pseudo-header (which must be valid UTF-8).

In the default ResolverBuilder::default_authority implementation, the path is now converted to a string using String::from_utf8_lossy. To prevent potential data loss, resolver implementations can override this trait method to handle invalid UTF-8 sequences as needed.