Dns telemetry should include time spent in async-over-sync queue

[MihaZupan]

In 6.0, #49171 added a per-host queue for async-over-sync Dns requests.
With that, the timing of ResolutionStart/Stop EventSource events changed.

For all other requests, the behavior is such that Start/Stop effectively wrap the entire time spent inside a Dns.Foo call.
For requests that hit the new queue, Start will only fire once the request is dequeued.
In E2E scenarios, this may manifest as a longer delay between the Http's RequestStart and ResolutionStart events, with no hint that the time was actually spent waiting inside name resolution.

We should modify the telemetry instrumentation such that the events also include the time taken on the queue.

Tagging subscribers to this area: @dotnet/ncl
See info in area-owners.md if you want to be subscribed.

Issue Details

---

In 6.0, #49171 added a per-host queue for async-over-sync Dns requests.
With that, the timing of ResolutionStart/Stop EventSource events changed.

For all other requests, the behavior is such that Start/Stop effectively wrap the entire time spent inside a Dns.Foo call.
For requests that hit the new queue, Start will only fire once the request is dequeued.
In E2E scenarios, this may manifest as a longer delay between the Http's RequestStart and ResolutionStart events, with no hint that the time was actually spent waiting inside name resolution.

We should modify the telemetry instrumentation such that the events also include the time taken on the queue.

Author:	MihaZupan
Assignees:	-
Labels:	area-System.Net, bug
Milestone:	-

[karelz]

Triage: We agree that it is useful thing to do. Moving to 6.0 as it is new feature/change adjustment.