[Feature] Multi-agent support in a single ACP bridge process

[HCH-hash]

Summary

Allow one paca-acp-bridge process to serve multiple agents instead of requiring one systemd service (or process) per agent.

Problem

paca-acp-bridge --agent-id currently accepts a single agent. A team of N ACP agents therefore needs N separate bridge processes/services, which complicates operations, monitoring, and resource usage.

Proposed Direction

• Extend the bridge CLI and configuration to accept multiple agent IDs (or a list) in one process.
• Multiplex the WebSocket / session handling so each agent retains its own identity, token, and workspace semantics.
• Keep the single-agent mode as the simple default for small setups.

Scope

• Product behavior
• Repository structure
• Architecture
• Developer experience
• Documentation
• Other

Additional Context

• Current limitation confirmed in apps/acp-bridge (single --agent-id / PACA_ACP_AGENT_ID).
• Large operational win for any team running more than a couple of ACP agents.
• Trade-off: process becomes longer-lived and slightly more complex; failure isolation between agents needs careful design.

[pullfrog]

  ^{｜ Build this ➔ ｜ Make a plan ➔}