Redesign error handling in the decode module

[partim]

This PR changes the entire way that errors are handled while decoding.

Sources now provide errors related to them failing the get more data when there should be more via the associated Source::Error type. This has been renamed from Source::Err for consistency with other such associated types.

A new type ContentError is introduced for errors where incorrectly encoded data is encountered, be actual encoding errors or data that isn't following the ASN.1 definition or informal profiles. This type wraps an error message that currently can be either a static str or a boxed Display trait object but can be extended if necessary later.

Since both source and content errors can happen during decoding, the compound type DecodeError wraps both these types. For content errors it also stores where in source the error happened to facilitate debugging. Currently, this type only allows displaying the error. It can be extended if additional access to the internally stored error is necessary.

The various content decoding methods now return a DecodeError.

In order to implement all these changes, the Source trait had to be adjusted. First, it needs to be able to provide the current position. This meant that it couldn’t be implemented on Bytes and &[u8] directly anymore. Therefore, the new trait IntoSource allows to convert a type into its Source implementation.

Finally, the trait's methods got cleaned up a bit. Specifically, Source::advance now only allows advancing as far as the length most recently returned by Source::request which also means it cannot fail anymore but needs to panic if the length is too large. This consistent to how Source::bytes already behaves.

[timbru]

LGTM

The error structure, here and in rpki-rs, looks quite elegant and flexible. I tried checking the actual message text used as well. I did not spot any issues - but if there is a need to rephrase things then I believe we can always do so later.