[feat] Add support for Pi Coding Agent (earendil-works/pi)

Problem

Follow-up to #550, where @elvizlai asked about pi-agent support and @jeff-r2026 suggested filing it as its own issue.

Some team members already use Pi (earendil-works/pi, npm @earendil-works/pi-coding-agent) as their daily coding agent, but TeamAI CLI doesn't recognize it yet. As a result:

• skills / rules / docs / hooks in .teamai/ can't be synced into Pi's native config directory via teamai pull;
• updates still have to be copied by hand into ~/.pi/agent/;
• team best practices built up around Pi can't be captured as a reusable team asset.

Proposed Solution

Add Pi as a new agent provider/adapter. Verified against Pi's own docs (settings.md, skills.md, extensions.md), its native resource paths are:

Notes:

• docs maps directly onto the existing .teamai/ → AGENTS.md distribution path already used for other agents — no special-casing needed.
• skills follow the standard Agent Skills spec (<name>/SKILL.md, description required).
• hooks are TypeScript "extensions" using pi.on(eventName, handler) (events like session_start, before_agent_start, tool_execution_end, agent_settled) — the hook dispatcher needs a Pi-specific event-name mapping.
• Pi has no built-in MCP config or custom sub-agent format (an explicit upstream design choice), so Phase 1 should cover skills / rules / docs / hooks only.

Alternatives Considered

• Manually pointing Pi's settings.json skills array at ~/.claude/skills: only covers skills, requires every user to hand-edit their own config, isn't automated by teamai pull, and doesn't cover docs/hooks — not a real substitute for a proper adapter.

Additional Context

• Upstream repo: https://github.com/earendil-works/pi (npm @earendil-works/pi-coding-agent, pi.dev)
• Pi's config layout follows the same agent-scoped skills/docs/hooks shape TeamAI already handles for other tools, so this should mainly be a matter of adding a new provider (path resolution + hook event mapping), not new architecture.