Use snafu to derive errors

[fjarri]

Removes the need of making separate Display impls for errors. Would be nice to use thiserror instead, but it doesn't support no-std.

Problems:

• It doesn't support enum variants with anonymous fields. The source error field must be named source and thus have a docstring. Annoying.
• Some errors cannot be wrapped (e.g. sec1::Error, which is returned by EncodedPoint::from_bytes()) because of trait bounds not satisfied.
• The strange "context" thing that it does where #[derive(Snafu)] enum MyError { Something } implicitly adds SomethingSnafu to the namespace, which is supposed to be used with the ResultExt trait.

So I don't know. Would be nice to remove some boilerplate, but I don't love this crate. Leaving the PR in the suspended state for now.

[codecov-commenter]

Codecov Report

Merging #113 (ffa12dc) into master (50a7a6e) will increase coverage by 1.34%.
The diff coverage is 0.00%.

❗ Current head ffa12dc differs from pull request most recent head 0763d32. Consider uploading reports for the commit 0763d32 to get more accurate results

@@            Coverage Diff             @@
##           master     #113      +/-   ##
==========================================
+ Coverage   47.49%   48.83%   +1.34%     
==========================================
  Files          17       17              
  Lines        1017      989      -28     
==========================================
  Hits          483      483              
+ Misses        534      506      -28     

Impacted Files	Coverage Δ
umbral-pre/src/capsule.rs	94.93% <ø> (+6.70%)	⬆️
umbral-pre/src/capsule_frag.rs	85.00% <ø> (+5.31%)	⬆️
umbral-pre/src/dem.rs	93.54% <ø> (+15.17%)	⬆️
umbral-pre/src/key_frag.rs	90.80% <ø> (+5.85%)	⬆️
umbral-pre/src/pre.rs	96.66% <0.00%> (+11.37%)	⬆️
umbral-pre/src/traits.rs	0.00% <ø> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.