Add support for telemetry

[ob-stripe]

r? @brandur-stripe @jameshageman-stripe

Add support for telemetry. The implementation is fairly similar to stripe-java's: stripe/stripe-java#661.

[jameshageman-stripe]

Thanks for doing this @ob-stripe! As far as I can tell the implementation looks good, and the tests are fantastic.

I added one small note about the concurrent test, but it really is a nitpick more than anything else.

[ob-stripe]

ptal @jameshageman-stripe

[jameshageman-stripe]

The pr tests indicate that both of the two requests sent in the concurrent test that have telemetry are both reporting req_1:

HttpClientHandler.SendAsync(Method: GET, RequestUri: 'http://localhost:12111/v1/balance', Version: 1.1, Content: <null>, Headers:
{
  User-Agent: Stripe/v1
  User-Agent: .NetBindings/22.9.0
  Authorization: Bearer sk_test_123
  X-Stripe-Client-User-Agent: {"bindings_version":"22.9.0","lang":".net","publisher":"stripe","lang_version":".NET Core 4.6.25211.01","os_version":"Microsoft Windows 10.0.14393 "}
  Stripe-Version: 2018-11-08
  X-Stripe-Client-Telemetry: {"last_request_metrics":{"request_id":"req_1","request_duration_ms":21}}
}, CancellationToken)
HttpClientHandler.SendAsync(Method: GET, RequestUri: 'http://localhost:12111/v1/balance', Version: 1.1, Content: <null>, Headers:
{
  User-Agent: Stripe/v1
  User-Agent: .NetBindings/22.9.0
  Authorization: Bearer sk_test_123
  X-Stripe-Client-User-Agent: {"bindings_version":"22.9.0","lang":".net","publisher":"stripe","lang_version":".NET Core 4.6.25211.01","os_version":"Microsoft Windows 10.0.14393 "}
  Stripe-Version: 2018-11-08
  X-Stripe-Client-Telemetry: {"last_request_metrics":{"request_id":"req_1","request_duration_ms":22}}
}, CancellationToken)

Don't suspect that this.prevRequestMetrics.TryDequeue(out requestMetrics) is the issue, so maybe theres another subtler problem with the test?

ptal @ob-stripe

[ob-stripe]

Ah yes, I think there's a potential race condition in the test itself. Incrementing invocationCount and generating the request ID string isn't an atomic operation, so two concurrent requests may end up with the same request ID. I'll try to fix this.

[ob-stripe]

Okay, I've factorized the code for setting up the requests in a FakeServer class and added a lock when incrementing the counter and generating the request ID string to fix the race condition.

ptal @jameshageman-stripe

[jameshageman-stripe]

@ob-stripe LGTM, thanks!