add support for full_refresh in template_fields

[mvdh98]

Summary

• Adds full_refresh to template_fields on DbtRunOperator, DbtSeedOperator, DbtCompileOperator, and DbtBuildOperator, enabling Jinja templating in DAG definitions.
• Adds string-to-bool coercion in TableMutabilityConfig.__post_init__ to handle Airflow rendering full_refresh as a string after template resolution.

Example usage

DbtRunOperator(                        
    task_id="dbt_run",                                                                                                                                                                                  
    full_refresh="{{ params.full_refresh }}",
)

Test plan

• Added test_full_refresh_in_template_fields for each of the 4 operators
• Added test_full_refresh_templated for each of the 4 operators
• Added test_table_mutability_config_full_refresh_string_coercion for string-to-bool coercion
• Verified existing tests still pass

Development Approach

• First added the tests and verified they failed
• Then made the code changes
• Then verified the tests now succeed and any already existing tests still succeed

[mvdh98]

@tomasfarias could you have a look at this one? :)

[tomasfarias]

@mvdh98 Thanks for the PR! I will do a more thorough review later, but for now I'll let CI run the tests.

[tomasfarias]

Pretty good work! I have a few questions, but overall great, shouldn't take too much more work to get it merged.

[tomasfarias]

A bit surprising that the tests use a string here. I don't think there is necessarily any issue with it, but I'm curious why didn't you use a bool? (i.e. assert op.full_refresh == True)

[tomasfarias]

It's a good test, but a tad repetitive, do you mind re-writing it using pytest.mark.parametrize? I can also push a quick commit for it.

[tomasfarias]

I wonder if it is safer to raise an error if the value is a str different than true/false. With this implementation, any typos (e.g. "treu") will be resolved to full_refresh=False, which I guess is the safer option as full_refresh=True is more destructive. But I think we should at least raise a warning.

Any thoughts on this?

[tomasfarias]

CI seems a bit flaky today, but the failures do not seem to be related to this PR. Retrying seems to help.

[tomasfarias]

@mvdh98 One final question: May I ask for the motivation for this change? Just curious as to know in case there is more work for us to do here. Did you encounter a situation in which this feature would have been useful?

Thanks!