[dotnet-watch tests] Cap output-wait timeout below the MTP hang-dump

[Evangelink]

Summary

Follow-up to the CI analysis on #55297. Fixes the test-infra half of the intermittent dotnet-watch.Tests hangs that kill an entire Helix work item (exit code 7) instead of failing a single test.

Problem

AwaitableProcess (the dotnet-watch test harness) derived its per-output-wait timeout directly from the Helix work-item timeout:

s_timeout = HELIX_WORK_ITEM_TIMEOUT is set ? (workItemTimeout - 10s) : 10 min;

With TestWorkItemTimeout = 60min (test/UnitTests.proj), that is 59:50. But Microsoft.Testing.Platform''s hang-dump fires at ~80% of the work-item timeout (48:00). So the hang-dump always wins the race against the graceful per-wait Assert.Fail("Output not found within …").

Result: when a single test wedges on an output wait (e.g. the intermittent WebSocket hot-reload connection hang in HotReload_WithWebSocketCapability — see #55258, #55044, #53289), the whole work item hangs ~48 min until the hang-dump, is killed with exit code 7, and loses all results — instead of that one test failing fast with a useful message.

Fix

Cap the per-wait timeout at 10 minutes (still honoring a shorter work-item timeout when configured), well below the 48-min hang-dump. A single output wait never legitimately needs more than a few minutes, so a genuinely wedged wait now fails that test fast with Output not found within 00:10:00, and the rest of the work item''s results are preserved.

This does not fix the underlying WebSocket connection flakiness (that''s a product-side change in the hot-reload agent — raise/retry the 5s connect timeout and surface the failure; deferred pending owner input). It only ensures the flake fails gracefully instead of nuking the work item.

Validation

• Microsoft.DotNet.HotReload.Test.Utilities builds clean (0 warnings / 0 errors).
• Change is confined to test infrastructure; no product code affected.

cc @tmat

[azure-pipelines]

Azure Pipelines:
Successfully started running 1 pipeline(s).
2 pipeline(s) were filtered out due to trigger conditions.
There may be pipelines that require an authorized user to comment /azp run to run.

[tmat]

hang that kills the whole run and loses all results

This is a bug (or lack of feature) in the test runner. The test outputs should be available no matter what.

[Evangelink]

@tmat I am taking that on me, I'll see how to fix/update MSTest or MTP depending on where the issue is.