Add include_traceback to model-backed retry policies

[kaxil]

LLMRetryPolicy and ClassifierRetryPolicy send the model only ExceptionType: message. That drops the exception chain, the stack and the module-qualified class name, which are often what says whether another attempt can succeed. A JSONDecodeError raised while handling a truncated read looks like bad data, so a model may say "stop", though the chained transport error underneath makes it a network failure a retry can fix.

include_traceback=True sends the output of traceback.format_exception() instead. The default is unchanged.

ClassifierRetryPolicy(
    llm_conn_id="jev_default",
    categories=CATEGORIES,
    include_traceback=True,
)

Design points:

• Redact first, then truncate, on the whole formatted traceback, so a registered secret can't be cut in half. That's the same order the message path already uses.
• A long traceback keeps its tail (with a leading (truncated) ... marker): the innermost frames and the final exception line say the most. The message path still keeps the head.
• Opt-in, because it sends more to the model: messages of chained exceptions, file paths on the worker, the source line of each frame, and more input tokens per classification. The docs warning now says so. Frame local variables are never included.

The single-argument traceback.format_exception(exc) form needs Python 3.10, which is the provider's floor.

---

• Read the Pull Request Guidelines for more information. Note: commit author/co-author name and email in commits become permanently public when merged.
• For fundamental code changes, an Airflow Improvement Proposal (AIP) is needed.
• When adding dependency, check compliance with the ASF 3rd Party License Policy.
• For significant user-facing changes create newsfragment: {pr_number}.significant.rst, in airflow-core/newsfragments. You can add this file in a follow-up commit after the PR is created so you know the PR number.