[Feature] Add persisted goal mode for long-running agent tasks

[yhw5231]

Affected area

Desktop UI (agent-gui / React)

Problem statement (what should this solve)

Long-running coding tasks currently require the user to manually send follow-up prompts after each model turn. Progress, pause/resume state, token usage, and recovery state are not persisted as a first-class conversation capability.

Proposed behavior

Add an optional per-conversation goal mode. Users can create and manage a persisted goal, see progress and TodoWrite steps above the composer, pause or resume work, recover the goal after reopening the conversation, and allow Agent-mode turns to continue automatically while the goal remains active. The run must stop on an explicit terminal state, budget or provider/runtime limits, and repeated no-progress output.

Estimated change scope

• crates/agent-gui/src/lib/chat/
• crates/agent-gui/src/lib/tools/
• crates/agent-gui/src/pages/chat/
• crates/agent-gui/src-tauri/src/commands/history/chat_history/
• Goal-mode regression tests and user documentation

Alternatives considered

Manual follow-up prompts do not preserve lifecycle state or progress across restarts. A global background worker would be less predictable and would not keep execution scoped to a conversation.

Pre-submit checklist

• I searched existing issues and pull requests and found no duplicates.
• This proposal is focused on a single feature or improvement.
• I understand a PR should come after this issue is confirmed by maintainers, otherwise it will be converted to draft.