feat: add distinguishable automation conversation titles

[Linxiushen]

Summary

• give prompt and plugin preset runs deterministic conversation titles before the first user message is sent
• include the automation name, compact trigger context, and a per-run identifier so repeated scheduled runs remain distinguishable
• extract GitHub/GitLab repository and issue/PR context when available
• keep title updates best-effort so metadata failures never abort the automation itself

Implementation

The dispatcher now includes the run ID in AUTOMATION_EVENT_PAYLOAD. Both preset tarballs include a shared title helper and PATCH the agent-server's existing conversation metadata endpoint immediately after conversation creation. Titles are normalized and capped at the server's 200-character limit while preserving the unique run suffix.

Examples:

Nightly repository review | cron 0 2 * * * | 12345678-123
PR review | github OpenHands/automation#274 | abcdef01-234

Validation

• 7 focused title tests passed
• Ruff check and format check passed
• Pyright: 0 errors, 0 warnings
• Python compilation passed for all changed Python files
• preset archive tests assert the shared helper is packaged for both prompt and plugin presets and invoked before send_message()

The full dependency suite was not run locally because the pinned LiteLLM build requires an MSVC linker in this Windows environment; repository CI provides the full Linux test run.

Closes #274

Disclosure: This contribution was prepared with AI assistance and independently reviewed and tested locally before publication.