bump version to 0.21.3

[apoelstra]

We've got a ton of minor changes in, plus fixing the Parity type and adding some extra serde impls. Let's push a minor version out so that we can move on to updating the upstream libsecp.

[tcharding]

For what its worth, LGTM. Although I'm not sure its worth having links to all those docs changes, they are all pretty trivial.

I checked the markdown renders correctly (on GitHub) and that all the links work.

[apoelstra]

Thanks! Normally I wouldn't bother linking to all the docs stuff, but there were a bunch and I could make the links pretty small.

[sanket1729]

utACK 61b6add. Looks good to me. Did not verify the release notes.

[apoelstra]

cc @elichai @TheBlueMatt could one of you ack this so I can merge and publish?

[Kixunil]

Note that the Parity change is theoretically breaking change because someone could write this:

let Parity { .. } = some_fn_returning_parity();

(or in other pattern matching)

Which works before and breaks after changing to enum. I know, it's dumb, I posted about this in IRLO recently. My personal approach for my libs is "doing this is too weird and hopefully unlikely and trivial to fix, if you do this sorry-not sorry, I'm making non-breaking release".

While I'm in favor of ignoring this I believe it's the right thing for me to point this out as others may have a different view.

[Kixunil]

One more thing, would you be up for ninja-doc-improvement of Parity before release? I now noticed the conversions number -> Parity are totally unclear - what values mean what? Could some panic?

[apoelstra]

@Kixunil sure, go for it.

[Kixunil]

Done. While working on it I noticed a few other problems:

• Parity is serialized as i32 - this may waste space in binary encodings. I think this is significant, do we want this?
• The serde error message is prefixed with "Expecting" which is wrong.
• It'd be nicer to map parity error as invalid_value in the Deserialize impl.

[apoelstra]

Oof, we should definitely fix the parity to be serialized as a single byte. I guess we can't do that in a minor release though.

The error stuff I think is ok to fix in a minor release, if you are so inclined.

But bear in mind that my intention is to do a major release basically ASAP after I get this merged, and to use the major release in the next rust-bitcoin release. Mainly I want #384 but I'm happy to mess with the Parity API too.

[Kixunil]

OK, will try to be quick and submit two PRs.

[Kixunil]

I found one more thing: Error doesn't carry the original value which can be annoying for debugging. Also it's not great that it's the same Error type as others while it can only return one variant. (Same stuff I want to solve at bitcoin.)

[sanket1729]

@Kixunil @apoelstra, anything I can help in unblocking this?

[Kixunil]

@sanket1729 I'm done with my PRs to this one.

[elichai]

ACK 61b6add

[apoelstra]

We need #407 first. Sorry, my bad.

In the meantime I'll update the changelog here assuming that we have 407.

[apoelstra]

@real-or-random @elichai can I get an ACK on this?

[elichai]

ACK