Fix race condition in ProcessBufferedRenderBatches_WritesRenders

[lewing]

Problem

ProcessBufferedRenderBatches_WritesRenders fails intermittently in CI (#61807), quarantined since Sep 2025.

Assert.True() Failure
Expected: True
Actual:   False
  at RemoteRendererTest.TestComponent.TriggerRender() in RemoteRendererTest.cs:line 778

Root Cause

The test had a ManualResetEventSlim that was never signaled — it was created, reset, and waited on with a 10-second timeout, but nothing ever called Set(). The comment acknowledged that "continuations of SetResult will not execute synchronously" but the synchronization mechanism was broken.

After firstBatchTCS.SetResult(), the renderer processes the batch completion asynchronously on its dispatcher. The test proceeded after the 10s timeout expired, then called TriggerRender() which asserts the render task completes synchronously. On loaded CI agents, the renderer's dispatcher was still busy processing the previous batch's continuation, causing the assertion to fail.

Fix

Move the post-SetResult() test setup (SetDisconnected + TriggerRender calls) inside a renderer.Dispatcher.InvokeAsync callback. This ensures:

1. All prior dispatcher work (from SetResult continuations) drains before our callback executes
2. TriggerRender runs with CheckAccess() == true, bypassing the internal task queue and executing the render synchronously

A single await InvokeAsync(() => {}) no-op was insufficient because InvokeAsync(Action) returns a task that completes inside PostAsync's callback, but PostAsync's own task (stored as _taskQueue) completes slightly later — creating a race where the next TriggerRender would chain instead of running synchronously.

Related to #61807

[Copilot]

Pull request overview

This PR re-enables and stabilizes the previously quarantined ProcessBufferedRenderBatches_WritesRenders test by changing how it waits for the renderer to finish processing the first render batch, aiming to eliminate intermittent CI failures in the Components Server circuit rendering tests.

Changes:

• Removes the [QuarantinedTest] attribute from ProcessBufferedRenderBatches_WritesRenders so it runs in CI again.
• Replaces an unused ManualResetEventSlim wait with a timeout-bounded loop intended to wait for render-batch processing to complete.

---

You can also share your feedback on Copilot code review. Take the survey.

[Copilot]

Pull request overview

This PR fixes an intermittent CI failure in the Components Server RemoteRendererTest by replacing a broken synchronization mechanism with a dispatcher-queued no-op, and re-enables the previously quarantined test.

Changes:

• Remove the quarantine marker from ProcessBufferedRenderBatches_WritesRenders so it runs in CI again.
• Replace the unused/unsignaled ManualResetEventSlim wait with await renderer.Dispatcher.InvokeAsync(() => { }) to correctly sequence dispatcher work after SetResult().

---

You can also share your feedback on Copilot code review. Take the survey.

[ilonatommy]

The fix looks fine.

Also removes the quarantine attribute so the test runs in CI again.

The process is to keep it in quarantined. We should merge this PR and add test-fixed label to the original issue (that we keep open - remove fixes keyword from the PR description). During the CI duty these tests that have been passing for a month will get moved from the quarantine.

feedback adressed

[ilonatommy]

Fun fact, copilot decided to add "fixes" keyword to squash message anyway because it s created by concatenating all commit messages of the PR and the 1st commit contained it.