[H/3] Fix code passed into QuicConnection.CloseAsync and QuicStream.Abort

[ManickaP]

Fixes the code passed into QuicConnection.CloseAsync and QuicStream.Abort which was by default -1. Quic supports only error codes up to 62 bits non-negative integer values: https://www.rfc-editor.org/rfc/rfc9000.html#integer-encoding.
Also added test which crashes the process (if running with DEBUG build of MsQuic) before this change and runs to completion afterwards.

Let me know if you want me to shuffle the code around, I'm not much familiar with the structure of your code base.

cc @amcasey

Fixes #55196

Also this will be followed up in S.N.Quic with some checks (dotnet/runtime#101385).

[amcasey]

I think it would be nice to flesh out the test a little, but I'm okay merging as-is if this unblocks you.

[amcasey]

My new check is failing CI. The test fails for me locally, but not until after the log message check succeeds, so I'm not sure why that's happening. I'll investigate.

Edit: it was flaky before my change.

[amcasey]

The windows failure is known CI flakiness

[amcasey]

Okay, looking more closely, the failure I'm seeing here is the same one I'm seeing locally and isn't related to the logging check. I think the test is flaky in that it won't reliably cause the connection to be closed on the server side.

[amcasey]

I believe the PR is ready to go.

[amcasey]

Ah, it was auto-merge from the previous design and my sign-off counts because it's technically Mana's PR. Well, shout if you don't like the exceptions @halter73.