Feat: Stop runtimes rather than delete them

[tofarr]

End-user friendly description of the problem this fixes or functionality that this introduces

Improved container management: Runtime containers are now stopped instead of deleted when not in use, allowing for faster reuse and better resource management. Containers are only deleted when a conversation is explicitly deleted.

• Include this change in the Release Notes. If checked, you must provide an end-user friendly description for your change below

---

Give a summary of what the PR does, explaining any non-trivial design decisions

This change improves how OpenHands handles runtime containers by stopping them instead of deleting them when they are no longer needed. This allows for faster container reuse and better resource management.

This PR changes how runtime containers are handled in OpenHands:

1. Changed default behavior to not keep runtimes alive by default (keep_runtime_alive=False)
2. Modified container cleanup to stop containers instead of removing them
3. Added proper container deletion when a conversation is deleted through the API
4. Improved container port handling and reattachment logic
5. Added new delete() class method to Runtime base class for explicit container cleanup

Design decisions:

• Stop instead of remove: This allows for faster container reuse since we don't need to recreate and pull images
• Default to not keeping runtimes alive: Better resource management by default
• Explicit deletion: Only delete containers when a conversation is explicitly deleted, giving more control to users

---

Screenshots

2 Stopped Conversations

Restart a conversation

Link of any specific issues this addresses

N/A

---

To run this PR locally, use the following command:

docker run -it --rm   -p 3000:3000   -v /var/run/docker.sock:/var/run/docker.sock   --add-host host.docker.internal:host-gateway   -e SANDBOX_RUNTIME_CONTAINER_IMAGE=docker.all-hands.dev/all-hands-ai/runtime:9172ca8-nikolaik   --name openhands-app-9172ca8   docker.all-hands.dev/all-hands-ai/openhands:9172ca8

[tofarr]

I had to change this logic because the existing logic did not work for stopped containers.

[tofarr]

Explicitly allow the runtime to run delete logic here.

[rbren]

Two issues I'm noticing:

• When I start a fourth conversation, the newest docker instance goes down, instead of the oldest
• When I resume an old conversation, it doesn't seem to respect the max of 3 running docker instances

[tofarr]

Two issues I'm noticing:

• When I start a fourth conversation, the newest docker instance goes down, instead of the oldest
• When I resume an old conversation, it doesn't seem to respect the max of 3 running docker instances

As noted on the call yesterday, the oldest conversation should now be the one that gets stopped, and the limitation of 3 instances is in place, but the conversation panel does not update in real time.