Make Options::apply fallible

[BenWiederhake]

This PR shows one way we could make Options::apply fallible.

This feature trivially enables error prioritization, e.g. raising different errors for --a=invalid --b=invalid and --b=invalid --a=invalid, which is what the GNU tools do.

Closes #112

[BenWiederhake]

Changes since initial push:

• Add date example, since this usecase motivated this change.
• Fix a test in shuf that exposes the same problem I noticed while writing date test: --help exits the entire test runner, thus hiding potential test failures. This is a workaround that needs a proper fix later.
• Something with clippy changed, it now complains about code I didn't even touch. So I fixed it by adding #[allow(unused)] to a handful of things.

[BenWiederhake]

Changes since last push: cargo fmt 😅

[BenWiederhake]

Changes since last push:

• Rebased
• Resolved trivial conflict (both this PR and Implement and exhaustively test cksum's weird implicit flags in 12 lines #122 added a test module)

[BenWiederhake]

Changes since last push:

• Resolved another trivial conflict (forgot to add the return-type in Settings::apply for the cksum-test)
• Fixed the date value-prefix logic. (We should make some nicer logic than this, but it's out-of-scope for this PR.)

[tertsdiepraam]

Awesome! My only real issue is with the Value thing I commented on. Once that's resolved: LGTM!

[tertsdiepraam]

I thought I built this into Value already? It's working here for example:

uutils-args/tests/coreutils/ls.rs

Line 480 in a3ecc62

let (s, _operands) = Settings::default().parse(["ls", "--format=acr"]).unwrap();

[tertsdiepraam]

So, I take a bit of issue with this example, because I don't like how Unspecified is a special thing that only exists for parsing. Essentially, it comes down to this: #11.

BUT! I don't care enough to block this. We can improve this later and we should move towards shipping this crate.