refactor: Replace pexpect with libtmux in BashSession

[xingyaoww]

This PR replaces the pexpect-based shell interaction with a more robust libtmux implementation, providing better reliability and functionality for command execution in OpenHands.

Key Improvements

• Replaced pexpect with libtmux for more reliable shell interaction
• Added proper handling of command errors, interactive commands, and timeouts
• Improved output capture and error detection
• Added automatic resource scaling for retries during evaluation
• Added comprehensive test suite to verify behavior

Technical Details

• Removed keep_prompt parameter from CmdRunAction as it's no longer needed with tmux
• Added tmux installation to all CI workflows (Ubuntu and macOS)
• Updated backend architecture diagram to reflect the new implementation
• Enhanced error handling with better timeout management
• Added resource scaling for retries in evaluation (up to 4x resources for failed attempts)

End-user Benefits

• More reliable command execution and output capture
• Better handling of interactive commands
• Improved error detection and reporting
• Automatic retry with increased resources for failed commands

Development Notes

Collaborated with OpenHands:

• Write tests for PS1 parsing https://www.all-hands.dev/share?share_id=41932d06ced712b1112e880ad3d3ed757308205438e0a6c6fd86048f22a5511d

---

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:31bdc15-nikolaik   --name openhands-app-31bdc15   docker.all-hands.dev/all-hands-ai/openhands:31bdc15
  -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:752f341-nikolaik \
  --name openhands-app-752f341 \
  docker.all-hands.dev/all-hands-ai/openhands:752f341

[diwu-sf]

Wow, this is really happening! Thank you so much!
With tmux, I suspect it will be easier to support "multi-terminals" like what Devin has, and have higher fidelity terminal "screen" captures to render in frontend.

[xingyaoww]

Finally figured out the resolve rate degradation issue! Now we get comparable performance on SWE-Bench Verified: 238 (tmux, this PR) vs 241 (reproduce-commit from 53% verified) resolved.

I will clean up the code and try to get this finally merged :D

[xingyaoww]

Very excited to finally merge this almost 2-month-old PR with nearly 300 commits 😄

attach a screenshot showing this is working

[diwu-sf]

Congrats on landing this one!