perf(ecs): increase staging worker CPU

[suguanYang]

Summary

• Increase the staging worker task from 512 CPU units to 1024 CPU units based on the measured 2 jobs/minute load-test evidence.
• Retain 4096 MiB memory and WORKER_CONCURRENCY=10.
• Add a rendering contract that protects the evidence-selected staging CPU and memory settings.
• No production task definition, workflow, environment variable, migration, queue, storage, or application behavior changes are included.
• Related infrastructure plan: Ontos-AI/knowhere-api-infra#22

Verification

• uv run --group lint ruff check deploy/ecs/test_render_task_definitions.py deploy/ecs/render_task_definitions.py
• uv run pytest deploy/ecs/test_render_task_definitions.py apps/worker/tests/contract/test_worker_shutdown_contract.py -q — 10 passed
• Rendered the staging worker artifact and verified CPU 1024, memory 4096, and concurrency 10.
• git diff --check
• uv run pytest -q — 473 passed, 4 unrelated failures, 14 warnings. Two summary-builder tests attempted the external https://example.com/v1/chat/completions endpoint; two page-memory tests use stale mocks that reject api_key and api_url. These failures are outside this two-file capacity change.

Deployment Notes

• This PR targets main only and does not deploy the change.
• No environment variables, database migrations, queue changes, or storage changes.
• Staging promotion and deployment require separate approval after merge.
• The staging service remains unchanged by this PR creation; rollback is the prior 512-CPU task definition revision.

Checklist

• Tests were added or updated when behavior changed
• Public docs, examples, or OpenAPI contracts were updated when needed — not applicable
• Database migrations are idempotent and safe to deploy — not applicable
• Logs, errors, and validation paths avoid leaking secrets or user data
• The pull request description explains any breaking or user-visible change — no breaking or user-visible change