Use ByteVector#fromHex in CSRF Middleware

[danicheg]

An attendant PR to the merged #6984.
As it was previously, the method from scodec needs some performance tweaks.
It'd be lovely if @mpilquist could do a miracle once again 🙏🏻
But, this time the difference isn't that dramatic for real-world usage.

[info] DecodeHexBench.decodeHexHttp4s    1000  thrpt    5  154.654 ± 11.577  ops/ms
[info] DecodeHexBench.decodeHexHttp4s   10000  thrpt    5   13.714 ±  0.116  ops/ms
[info] DecodeHexBench.decodeHexHttp4s  100000  thrpt    5    1.524 ±  0.507  ops/ms
[info] DecodeHexBench.decodeHexScodec    1000  thrpt    5   81.527 ±  4.917  ops/ms
[info] DecodeHexBench.decodeHexScodec   10000  thrpt    5    3.923 ±  0.063  ops/ms
[info] DecodeHexBench.decodeHexScodec  100000  thrpt    5    0.425 ±  0.024  ops/ms

[armanbilge]

I'm just happy to see internal vanishing before my eyes!

Hard to reason about performance here, since Hmac cryptography is involved here, that might be the bottleneck anyway. Not sure tho, maybe that's fast 😇

[mpilquist]

I did some quick benchmarking of scodec's fromHex and it turns out there's a bunch of integer boxing as a result of using a pattern match on a Char. Working on a PR now.

[mpilquist]

Ignore that - the boxing was artifact of the profiler.

[mpilquist]

WIP here: scodec/scodec-bits#422

[danicheg]

scodec/scodec-bits#422 is merged, so we are unlocked here. Steward will bring us a new scodec version when it is published.