Make growing types non exhaustive

[faern]

I was about to bump the version and release. Given that we have added an error variant to an enum and a field to a struct it's technically a breaking change even if the feature is off by default... At least I think it should count as that 🤔 Because even if you as a direct user of udp-over-tcp does not activate the feature, a transitive dependency might....

The fix for the error enum is easy, make them #[non_exhaustive]. I think this is a pretty good default for error enums to begin with. Because it allows adding more error cases without it being a breaking change...

But what about the Options struct? I figured the best might be to also make it #[non_exhaustive] and provide facilities for users to construct it without hitting compilation errors both if others enable the statsd feature OR if we add other optional fields later.

We still need to bump the version in a breaking fashion after this PR (to 0.4.0) but these changes will allow us to stay on 0.4.x after that even if we add optional option fields or error variants, something we cannot if we don't merge this PR.

---

This change is 

[faern]

Reviewable status: 0 of 4 files reviewed, 2 unresolved discussions (waiting on @dlon)

---

src/tcp2udp.rs line 24 at r1 (raw file):

/// This struct is `non_exhaustive` in order to allow adding more optional fields without
/// being considered breaking changes. So you need to create an instance via [`Options::new`].
#[derive(Debug, Clone)]

The Clone part here is not important. I just felt like there is no reason for it not to be clonable, and it might make it more ergonomic to use.

---

src/tcp2udp.rs line 48 at r1 (raw file):

    /// Host to send statsd metrics to.
    #[cfg_attr(feature = "clap", clap(long))]
    pub statsd_host: Option<SocketAddr>,

Oops. Forgot to set this to pub in the previous PR. This was clearly a bug, since users of the library must be able to set this field.

[dlon]

Reviewed 4 of 4 files at r1, all commit messages.
Reviewable status: all files reviewed, 2 unresolved discussions