Improve f64 encoding

[vmx]

Copy the bytes into a buffer instead of destructuring them. On my machine it's an about 15% performance improvement.

A benchmark was added, which can be run via

cargo bench --bench f64_enc --features use_alloc

---

I've only optimized the f64 code path as this is what I discovered, when I was switching a project over to cbor4ii. If other encodings should be changed/benchmarked as well, let me know.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.14%. Comparing base (30fe080) to head (f3ece39).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master      #51      +/-   ##
==========================================
+ Coverage   85.82%   86.14%   +0.31%     
==========================================
  Files          11       11              
  Lines        1912     1920       +8     
==========================================
+ Hits         1641     1654      +13     
+ Misses        271      266       -5     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[quininer]

Interesting, I thought LLVM could optimize it. would you mind improve u64 as well?

https://godbolt.org/z/E5Pdx7jjT

[vmx]

I went all in and changed it for all numeric types, as even for 16-bit, there are less assembly instructions (though pobably not measurable: https://godbolt.org/z/ov97Pq43j

I did it for all as I'm a fan of code symmetry. If I should only do it for the 64-bit types, I'm happy to change it again.

I've also removed the benchmark, as it was for illustration purpose only and I don't think it makes sense to keep it around (less bit rot).

[quininer]

Thank you!

[vmx]

Thanks for merging.

I'm still looking into another encoding bottleneck. It could well be that it's not a cbor4ii thing, but related to other parts of the library. I just wanted to mention as you might want to wait for a new release until all the changes I'm currently working on are in.

[vmx]

The other bottleneck was something else. So don't expect any perf related PRs for now.

For anyone who's reading this and might be interested what the other performance things was. I was using the BufWriter from cbor4ii. That is slower than the one from the Rust standard library.