Add an Agent Canvas local and Cloud parallel-work use case

I can

I can follow a public example to parallelize work locally or on OpenHands Cloud from Agent Canvas, track each child, recover from failure, and review results before integration.

Scope

Add a concrete parallel-work page to the existing examples/use-cases documentation structure and link it from the relevant overview/navigation. Document the public local implementation and only the existing Cloud API contract.

Acceptance criteria

• The local walkthrough launches at least two isolated child conversations from a parent using the typed Canvas action.
• The Cloud walkthrough launches at least two children using the same user-level flow and the existing Cloud API.
• Both walkthroughs show dependency-aware task briefs, child status tracking, direct child opening, and parent reload.
• A failure/recovery example demonstrates retry or relaunch without losing successful sibling work.
• The integration step includes an explicit human review gate and verification of combined results.
• The page explains isolation, concurrency limits, shared-state hazards, and when not to parallelize.
• Screenshots or equivalent live evidence match the released UI and skill behavior.
• The page is linked from the docs' examples/use-cases entry point and passes docs checks.
• No private-repository implementation or new Cloud API is presented as a prerequisite.

Dependencies

• OpenHands/software-agent-sdk#4751 — aggregate clients.
• OpenHands/agent-canvas#1987 — shared Cloud transport.
• OpenHands/software-agent-sdk#4119 — local relationships.
• OpenHands/OpenHands#15461 — typed launch.
• OpenHands/OpenHands#15460 — tracking and controls.
• neubig/workflow#9 — delegation skill and examples.