[SPARK-58754][INFRA] Update JIRA fix versions for separately backported pull requests

[zhengruifeng]

What changes were proposed in this pull request?

Update dev/merge_spark_pr.py so that JIRA handling is not skipped when a backport is
cancelled after the original merge has already been pushed.

The script now also runs JIRA handling when an already-merged pull request is invoked again
in backport mode. If the associated issue is already resolved, the script preserves its
existing Fix Versions and offers to add only the versions inferred from newly discovered
backport branches.

Why are the changes needed?

The JIRA update previously ran after the optional cherry-pick loop. Cancelling a cherry-pick
exited through the loop's cleanup path and never reached the JIRA update. A later invocation
recognized the closed pull request as a backport, but that path also exited before JIRA
handling.

This happened with #57754: the first invocation merged into master and was cancelled while
starting the branch-4.x backport. The second invocation completed that backport, but neither
invocation updated SPARK-58556, so it had to be resolved manually with both Fix Versions.

Does this PR introduce any user-facing change?

No. This only changes the Spark committer merge tool.

How was this patch tested?

Added inline doctests for retaining existing Fix Versions while selecting only newly inferred
backport versions.

The following checks passed:

cd dev
conda run -n spark-dev-313 python -m doctest -v merge_spark_pr.py

79 doctests passed.

conda run -n spark-dev-313 ruff check --config pyproject.toml dev/merge_spark_pr.py
conda run -n spark-dev-313 ruff format --check --config pyproject.toml dev/merge_spark_pr.py

Was this patch authored or co-authored using generative AI tooling?

Generated-by: Codex (GPT-5)

[uros-b]

Thank you @zhengruifeng and @HyukjinKwon!