Improve crypto performance

[yaakov-h]

Benchmark:

Benchmark run against a packet about 670 bytes long.

Current code:

• Time to encrypt an outgoing packet: 6.464us
• Memory allocations to encrypt an outgoing packet: 6320B
• Time to decrypt an incoming packet: 6.524us
• Memory allocations to decrypt an incoming packet: 5208B

This branch:

• Time to encrypt an outgoing packet: 4.054us
• Memory allocations to encrypt an outgoing packet: 848B +size of packet
• Time to decrypt an incoming packet: 3.688us
• Memory allocations to decrypt an incoming packet: 1008B + size of packet

Unfortunately it seems that about 2/3rds of the CPU time is spent computing the HMAC. I don't think we can do anything about that at the moment.

See also: #552

Some discussion on previous PR at #554.

[codecov-io]

Codecov Report

Merging #683 into master will increase coverage by 2.66%.
The diff coverage is 0%.

@@            Coverage Diff             @@
##           master     #683      +/-   ##
==========================================
+ Coverage   22.07%   24.73%   +2.66%     
==========================================
  Files          86       90       +4     
  Lines        8666     9508     +842     
  Branches      714      850     +136     
==========================================
+ Hits         1913     2352     +439     
- Misses       6627     7022     +395     
- Partials      126      134       +8

Impacted Files	Coverage Δ
SteamKit2/SteamKit2/Util/CryptographicContext.cs	0% <0%> (ø)
...2/Networking/Steam3/NetFilterEncryptionWithHMAC.cs	0% <0%> (ø)	⬆️
...t2/SteamKit2/Steam/Handlers/SteamUser/Callbacks.cs	2.31% <0%> (-0.76%)	⬇️
SteamKit2/SteamKit2/Types/DepotManifest.cs	0% <0%> (ø)	⬆️
...mClient/Configuration/SteamConfigurationBuilder.cs	100% <0%> (ø)	⬆️
...SteamKit2/Networking/Steam3/WebSocketConnection.cs	0% <0%> (ø)	⬆️
...Kit2/SteamKit2/Util/SteamKitWebRequestException.cs	100% <0%> (ø)
SteamKit2/SteamKit2/Util/KeyValuePairExtensions.cs	100% <0%> (ø)
...2/SteamKit2/Steam/WebAPI/WebAPIRequestException.cs	100% <0%> (ø)
... and 20 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 49c61b6...b5b47ba. Read the comment docs.

[xPaw]

What is this blocked on?

[yaakov-h]

Merge conflicts, tests, and ideally I'd rather use Span<> than ArraySegment<> but that means either taking a package dependency or dropping <= netstandard2.1 I think, and I'm not sure which way we want to go on that overall.