Agent status mapping, parallel dashboard loading, rig parsing, bd exec fixes

[yougotborked]

Changes

• Richer agent status: proper state mapping, rig agent distinction, current_task tracking
• Dashboard: parallel data loading via Promise.allSettled for status/convoys/work/mail
• Server: robust rig list parsing with JSON fallback to text parsing
• BDGateway: use bd executable, keep --no-daemon compat, use create command
• Various UI/component improvements across sidebar, agent-grid, work-list, etc.

21 files changed, 603 insertions, 310 deletions

[web3dev1337]

This PR introduces regressions that should block merge. I opened a replacement PR with the fixes and full validation here: #28

Key issues in #27:

• websocket refresh handling now relies on state.currentView, but that field is never defined or updated, and several emitted events are no longer handled, so the UI can go stale until a manual refresh
• /api/agents and the dashboard assume rig.agents, but the current backend contract and tests still use legacy rig.hooks, which drops rig workers from summaries and agent listings
• BDGateway bypasses the resolved executable path and hardcodes bd, which breaks configured executable overrides and weakens the current trust boundary
• main refresh paths no longer force cache-busting even though the affected endpoints are still cached, so refreshes can legitimately return stale data

Replacement PR #28 keeps the useful parts of the submission, restores compatibility, and adds regression coverage.

[web3dev1337]

Thanks for the revision, @yougotborked. I reran the full suite and kept your authored commit (afeb7b0) preserved in the salvage path. We’re merging PR #28 via merge commit (not squash) so your contribution remains directly in project history.