Implement context-specific tags on #[derive(Sequence)]

[bstrie]

This also adds a context-specific tag test to the code for #[derive(Choice)], and cleans up the tests in these files.

If you'd like I would be happy to add more tests to cover the interactions between the various options, but I admit that I'm not an ASN.1 expert so I would need some guidance as to how these options interact so that I could formulate the correct desugaring.

Closes #322.

[npmccallum]

@bstrie I think you mean context specific, not context sensitive.

[tarcieri]

I think you mean context specific, not context sensitive

Thanks to the Chomsky Hierarchy I have made this mistake several times myself 😅

[npmccallum]

@bstrie I think you need to fix the commit message wording too.

[npmccallum]

@bstrie Also, one of the tests is failing.

[tarcieri]

One of the errors is on IssuingDistributionPointsExample:

    impl ::der::Sequence<'_> for IssuingDistributionPointExample {
        fn fields<F, T>(&self, f: F) -> ::der::Result<T>
        where
            F: FnOnce(&[&dyn der::Encodable]) -> ::der::Result<T>,
        {
            ()
        }
    }

...not sure what's happening there but the empty tuple is definitely not the right result.

Looks like something similar on the other error:

    impl<'a> ::der::Sequence<'a> for OneAsymmetricKey<'a> {
        fn fields<F, T>(&self, f: F) -> ::der::Result<T>
        where
            F: FnOnce(&[&dyn der::Encodable]) -> ::der::Result<T>,
        {
            ()
        }
    }

Perhaps the unparseable tokens are being lost?

error: proc-macro derive produced unparseable tokens

[bstrie]

Code updated to pass all the tests (sorry about not doing so the first time, I forgot to test with --all-features). The tests that were failing were already using context-specific tags (which I guess were just being ignored before now) in conjunction with various other field attributes. Taking into account all the supported combinations of the currently-implemented attributes was a bit tricky and required some subtle changes to make sure all the references lined up correctly, but in the end I think I managed to find a nice arrangement that doesn't involve adding state to the Lowerer or else compromising the unit-testability of the Lowerer methods. Once again I tested this with #323 to confirm that it addresses #322 .

[tarcieri]

Oh my 😅

[bstrie]

We call that "self-documenting code". :P