Allow Common AI SQL imports without DataFusion

[aaron-y-chen]

Summary

LLMSQLQueryOperator eagerly imports DataFusionEngine. This works in the development environment because apache-airflow-providers-common-sql[datafusion] is installed as a dev dependency:

airflow/providers/common/ai/pyproject.toml

Line 142 in fe9d865

"apache-airflow-providers-common-sql[datafusion]",

However, the Common AI sql extra does not install DataFusion, so importing the operator fails in production installations where DataFusion is absent.

Bug Reproduction

Use docker to install airflow and common-ai provider, then import LLMSQLQueryOperator.

$ docker run --rm -it python:3.12-slim bash
$ python -m venv /tmp/airflow-repro
$ . /tmp/airflow-repro/bin/activate
$ python -m pip install --upgrade pip

$ python -m pip install --no-cache-dir "apache-airflow==3.3.0"  "apache-airflow-providers-common-ai[sql]==0.6.0"

$ python -m pip show datafusion
WARNING: Package(s) not found: datafusion

$ python -c "from airflow.providers.common.ai.operators.llm_sql import LLMSQLQueryOperator"
ModuleNotFoundError: No module named 'datafusion'

After fix

Build the Common AI provider wheel from this PR, then install its sql extra with Airflow 3.3.0 in a fresh python:3.12-slim container.

$ python -m pip show datafusion
WARNING: Package(s) not found: datafusion

$ python -c "from airflow.providers.common.ai.operators.llm_sql import LLMSQLQueryOperator; print('operator import: OK')"
operator import: OK

$ python -c "from airflow.providers.common.ai.decorators.llm_sql import llm_sql_task; print('decorator import: OK')"
decorator import: OK

---

Was generative AI tooling used to co-author this PR?

• Yes (please specify the tool below)

Generated-by: [Claude Fable 5] following the guidelines

[kaxil]

@nailo2c Can you rebase on main please to resolve merge conflicts?

[aaron-y-chen]

Sure, let me do it.