Refactor implementation of Decode::skip and Decode::encoded_fixed_size

[kalaninja]

• add Input::skip
• refactor implementation of Decode::encoded_fixed_size
• refactor implementation of Decode::skip

By default, Input::skip does an actual read and discards the result, implementors should provide a specialized implementation.

By default, Decode::skip executes Input::skip for types providing Decode::encoded_fixed_size. Otherwise, it is just calling Decode::decode., so implementors should provide a specialized implementation.

As discussed in previous attempts that I found on github, Decode::skip shouldn't validate the input data, providing a fast forward functionality instead. IMHO if a calling code wants to check for data integrity they can manually do something like let _ = Decode::decode(input)?;

Closes #208
Closes #244

[kalaninja]

@ggwpez @bkchr guys, what do you think about this?

[gui1117]

Looks good.

When skipping we wouldn't do the descend_ref/ascend_ref. I don't if this is an issue.

I don't know if tests really test for skip, encoded_fixed_size.
Ideally we should have for every types, a bunch of value, with the expected encoded_fixed_size, then we test decode, skip and encoded_fixed_size.

[gui1117]

I added some tests on your branch

[gui1117]

Thinking about descend_ref/ascend_ref. it is mostly used with decode_with_depth_limit.

I think we can call them, or we can just have a comment on the skip specification saying, skip doesn't take into account descent_ref and ascend_ref.

But at a first glance implementing skip with descend_ref/ascend_ref seems not difficult so maybe we should just use them.

WDYT?

[kalaninja]

Yes, I agree descend_ref/ascend_ref should be used when skipping. I added a corresponding commit to enable this for implementors of WrapperTypeDecode and everything that is skipped (encoded) as Vec.

[gui1117]

Looks good to me, I added a minor test.