fix: support both ColabKernelClient and KernelClient in runtime

[Mhdaw]

Summary

Fixes an AttributeError when running colab run / colab exec caused by attribute differences between PyPI releases of jupyter-kernel-client (KernelClient) and the git repository fork (ColabKernelClient).

Root Cause

• PyPI release of jupyter-kernel-client exports KernelClient.
• Git repository fork of jupyter-kernel-client exports ColabKernelClient.
• When installed as a package via pip install google-colab-cli or uv tool install google-colab-cli, jupyter-kernel-client from PyPI is resolved, causing AttributeError: module 'jupyter_kernel_client' has no attribute 'KernelClient'.

Fix

Added dynamic attribute inspection in ColabRuntime.kernel_client to instantiate ColabKernelClient when present and fall back to KernelClient. Updated test_runtime.py unit tests accordingly.

Testing

• Executed full test suite (uv run pytest): 322/322 passed.
• Verified live script execution via colab run on Google Colab VM.

[google-cla]

Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA).

View this failed invocation of the CLA check for more information.

For the most up to date status, view the checks section at the bottom of the pull request.

[Mhdaw]

Context & Background

While setting up automated Google Colab workflows using Google Antigravity paired with Gemini 3.6 flash high, we encountered a runtime AttributeError when executing colab run headlessly:

AttributeError: module 'jupyter_kernel_client' has no attribute 'KernelClient'

Upon investigating the dependency resolution, Gemini identified the root cause: a structural difference between the official PyPI release (jupyter-kernel-client==1.0.0, which exports KernelClient) and the development git source (googlecolab/jupyter-kernel-client, which exports ColabKernelClient).

This PR adds dynamic attribute inspection in ColabRuntime to support both package sources seamlessly, guaranteeing compatibility for both standard pip/uv PyPI installations and source builds.

Verified against live Google Colab runtimes and passed the complete 322-test unit suite (uv run pytest).

[sethtroisi]

Thanks for the PR!