Incorrect error handling for NXDOMAIN

[LeifOfWilsonCreek]

When receiving the response NXDOMAIN, AsyncDNSResolver throws .internalError, requiring callers to code like this:

        var records: [SRVRecord]
        do {
            records = try await Self.resolver.querySRV(name: srvName)
        } catch let error as AsyncDNSResolver.Error where error.code == .internalError {
            // Catch NXDOMAIN, which is erroneously reported as `.internalError`.
            // NXDOMAIN and `results==[]` (commonly called "NODATA") both indicate affirmative nonexistence.
            // Hopefully other common errors aren't `.internalError` as that may cause weird behavior here.
            records = []
        } catch let error {
            // Catch timeouts and other transient network errors.
            throw error
        }

Adding an exception for .nxDomain might be a tempting solution, but it would break existing workarounds like the above. But it can't be conflated into .internalError; doing so as it does now is clearly a bug.

I suggest updating queryXXX calls to return [] when the response is coded NXDOMAIN. This makes sense because the distinction between the DNS entry existing only by different types (commonly "NODATA") vs not existing at all (NXDOMAIN) isn't a useful distinction, so they can be treated the same. Returning [] simplifies clients by not needing the first catch case shown above, and sidesteps potential compatibility issues for existing callers.

[Lukasa]

This sounds like a reasonable enhancement, and we'd welcome a PR that makes this change.