[FEATURE]: Native reload for long-running instances #19120

Feature hasn't been suggested before.

• I have verified this feature I'm about to request hasn't been suggested before.

Describe the enhancement you want to request

For a long-running opencode server, refreshing runtime state should not require restarting the whole server process.

In multi-client setups, a full restart is too disruptive because it can interrupt unrelated sessions. What is needed instead is an instance-scoped soft reload of the current workspace context.

I think reload should be a native core capability with two entry points:

• a native reload tool for agents
• a TUI reload command for users

Both should use the same core reload path.

The reason I think this belongs in core is that reload is a runtime/instance action, not a plugin-specific feature. Agents may need to refresh their system context after instance-level state changes or runtime reinitialization, and doing that by restarting the whole server is not realistic.

I already have a working implementation in a fork. The current approach is not a server restart: a reload request marks the instance as pending, active sessions keep running until they reach the next safe wait point in the loop, and once the last active session reaches that point the current instance is disposed and forgotten so the next access recreates it. Reload status is also published as an event so clients can react consistently.

This is only about soft reloading the current instance/workspace context, not binary self-update, package update, or full server restart.

This seems related to #6719, #8751, #18119, and #6815, but I think this is a slightly different proposal. Those issues are mostly about reloading config, hot-reloading content, or adding a TUI entry point. What I’m proposing is a native instance-scoped soft reload capability in core, exposed to both agents and the TUI.

If this direction matches the core design, I can prepare a small PR.