fix(sdk-ts): scope polling pause time to each wait - #1163

Summary

Reusing a parent PauseClock currently extends a new waitForExecutionResult timeout by all time the execution previously spent paused. For example, an earlier 4-second pause turns the next 200 ms polling budget into roughly 4.2 seconds.

Snapshot the cumulative paused time when each wait starts, then subtract only additional paused time. The shared execution clock, WAITING transitions, and status callbacks retain their existing behavior.

Type of change

• Bug fix

Test plan

• ./scripts/test-all.sh — control plane and Go SDK passed; Python SDK: 2,959 passed, 4 skipped, 38 deselected; TypeScript SDK: 1,392 passed, 1 skipped; Web UI: 837 passed.
• make fmt tidy — passed; unrelated formatting output from the existing baseline was reverted.
• cd sdk/typescript && npm run lint — passed.
• Client/pause regression tests — 46 passed. New cases use a real local HTTP server and controlled Date values to cover successful polls, HTTP 503 retries, and new WAITING intervals.

Go checks ran in the existing isolated Linux Go 1.25.14 environment with an offline public-DNS fixture; Python and Node checks ran locally. The suite's default exclusions were retained. The two past-pause cases fail on the original implementation and pass with this fix.

Test coverage

• I ran tests for the changed surface locally.
• New code paths are covered by regression tests.
• Coverage Summary is green in CI before requesting review.

CI coverage and hosted CLA checks remain to be evaluated after PR creation.

Checklist

• I read docs/CONTRIBUTING.md and docs/DEVELOPMENT.md.
• The commit is signed and follows conventional-commits style.