feat(agents): add max_agents_per_repo to allow multiple agents per repo

[PieterPel]

Closes #43

Summary

• Adds agents.max_agents_per_repo config field (default: 1, fully backward-compatible)
• Adds State::project_agent_count() to count running agents per project
• Refactors is_project_busy() to delegate to project_agent_count()
• Updates try_launch and auto_launch to enforce the configurable limit
• Improves status message to show N/MAX agents active per repo

Why

When git.use_worktrees = true, each ticket gets an isolated worktree so there's no conflict between concurrent agents on the same repo. The previous hard boolean is_project_busy() made the 1-per-repo limit impossible to lift even when worktrees made it safe to do so.

Usage

[agents]
max_parallel = 5
max_agents_per_repo = 3  # new — default is 1 (no behaviour change)

Test plan

• cargo check passes
• New unit tests for project_agent_count, max_agents_per_repo default, and is_project_busy consistency
• Existing tests unchanged

[untra]

👋 heya, thank for the contribution . I might include this in the next release, but with some adjustments.

I don't have the infrastructure to take community contributions in this project just yet, but I would like to help you out with your operator use-case . I'm trying to unify a number of different tools to define auditable AI workflows and standardizing governance controls for AI doing kanban shaped work. If this is something I can help you with, please reach out to me! email, X, linkedin DM happy to chat privately.

[untra]

🎻