[codex] Fix best practice resource update and lookup regressions

[07-clottattoos-cyber]

Summary

This PR fixes two regressions that currently break the official best_practice example workflow described in the documentation.

The affected workflow is the documented example that:

1. uses host_node/create_resource to create or update liquid in an existing container
2. uses workstation/PumpTransferProtocol to transfer that liquid into another container

In v0.10.19, the workflow can fail in two places:

• create_resource crashes when updating an existing RegularContainer
• PumpTransferProtocol can fail to resolve from_vessel=container during resource lookup

Closes #256.

Root cause

The first regression comes from an API mismatch during the container state refactor:

• RegularContainer now relies on serialize_state()/load_state()
• the update path in append_resource still directly accesses .state

The second regression comes from resource lookup being too bridge-dependent:

• the documented workflow passes a bare vessel id like container
• the remote lookup path does not always resolve that identifier correctly
• the local host already has enough resource context to resolve it safely

Changes

• make append_resource update container state through the current resource state API, while keeping compatibility with legacy .state
• resolve resources locally in host_node before falling back to bridge requests
• add a startup-config-based fallback for local resource resolution when needed by the running workflow context

Validation

Validated in an isolated environment against the official documented workflow:

• host_node/create_resource succeeds when targeting an existing container
• workstation/PumpTransferProtocol succeeds afterwards
• the end-to-end cloud-triggered workflow completes successfully

Additional local check:

• python -m py_compile unilabos/ros/nodes/base_device_node.py unilabos/ros/nodes/presets/host_node.py