fix(test): increase timeouts to reduce CI flakiness on slow runners

[mjameswh]

Fixes several integration tests that flake on slow GHA runners (Windows, macOS, linux-arm) due to tight timeouts:

• worker-connection-replacement: waitUntil after connection switch 10s → 20s
• root execution is exposed: waitUntil for child workflow start 8s → 20s
• sinks / Core issue 589: wf.condition timeout 10s → 20s; on slow linux-arm Bun runners the timer was firing before the test could send the unblock signal
• Lang's SDK flags replay correctly (Bun): wrap __temporal_workflow_metadata query in asyncRetry to handle transient DEADLINE_EXCEEDED timeouts that were triggering an unserializable CombinedWorkerRunError on Bun

[THardy98]

Activity reset/cancellation tests are also flaking
(not suggesting you tackle them in this PR, but just mentioning for tracking. Might be worth just bumping timeouts there too)

[mjameswh]

Two more CI failures occurrences of the following error:

integration-workflows › Activity reset without retry returns expected cancellation details
Difference (- actual, + expected):
- undefined
+ {
+   cancelRequested: false,
+   notFound: false,
+   paused: false,
+   reset: true,
+   timedOut: false,
+   workerShutdown: false,
+ }
› <anonymous> (src/test-integration-workflows.ts:1513:7)
› async <anonymous> (/Users/runner/work/sdk-typescript/sdk-typescript/packages/worker/src/worker.ts:1988:24)