[Feature Request] Add proactive context compaction before token limit with agent state preservation

Bug Description
It would be nice if it was always possible to /compact (main and agents) before running into a usage / token limit.
Or if it would automatically do that or warn before spawning a new agent.

Now when i hit the usage wall, all the work of agents (which may be 90% done) appears to be discarded AND i get a big penalty for reviving an old (many hours or sometimes days old) thread.

Environment Info

• Platform: darwin
• Terminal: Apple_Terminal
• Version: 2.1.173
• Feedback ID: 8c4b57a0-41ab-4c8e-a123-ab4d41837801