Allow infallible construction of Signature

[yukibtc]

Hi! This PR add 1 constructor and 2 methods to Signature struct:

• Add Signature::from_byte_array constructor
• Add Signature::to_byte_array and Signature::as_byte_array methods
• Deprecate Signature::serialize method

[Kixunil]

Wow, I haven't noticed that signatures actually don't have any invariants (other than length). Concept ACK though I'm not sure about the renaming of serialize.

[Kixunil]

Also the title is misleading since this is not what "zero copy" means. "Zero copy" is just a stupid name for "zero heap allocations" and the original code didn't do any heap allocations in the first place. Further, in hashes we use from_byte_array for array-accepting constructors.

What I believe is non-controversial about this PR: we want some infallible method that accepts an array. I wanted to suggest you to make just that change first but then I remembered about from_byte_array thing.

[apoelstra]

Neat! Concept ACK.

I would like to use to_byte_array and from_byte_array to be consistent with hashes. (I have proposed the same thing for SecretKey in #724.)

[yukibtc]

Also the title is misleading since this is not what "zero copy" means. "Zero copy" is just a stupid name for "zero heap allocations" and the original code didn't do any heap allocations in the first place

Ahh, thanks for the clarification

---

I've done the changes as suggested.

[Kixunil]

ACK 23b87a6

[Kixunil]

We usually use TBD in since and mass-replace it during release but there isn't anything wrong with this.

[apoelstra]

ACK 23b87a6 successfully ran local tests