[SPARK-60139][PYTHON] Remove dead code UNSUPPORTED_COMBINATIONS

[bearomorphism]

What changes were proposed in this pull request?

Remove the unused UNSUPPORTED_COMBINATIONS constant from python/pyspark/install.py, together with its import and dead loop in test_checked_versions.

This supersedes the original annotation-only approach. The existing supported-version checks and explicit test rejecting Hive 1.2 remain unchanged.

Why are the changes needed?

History investigation shows that this constant has been unused by production code since September 2020:

• September 23, 2020: SPARK-32017, commit 942f577b6e3 introduced the constant to reject Hive 1.2 paired with either without-hadoop or Hadoop 3.2.
• September 24, 2020: SPARK-32982, commit 688d016c7ac removed Hive 1.2 support, emptied the list, and deleted the production membership check in checked_versions. The empty declaration and test loop were left behind.
• September 27, 2020: SPARK-32714, commit c65b64552f9 added # type: ignore to silence mypy's missing-annotation error for the empty list.

Before this cleanup, a repository-wide search found only the declaration, the test import, and the test loop. The loop executes zero iterations because the list is empty. Removing these leftovers is preferable to adding a type annotation to dead code.

Does this PR introduce any user-facing change?

No.

How was this patch tested?

The following focused tests passed both before and after removal (2 tests):

source .venv/bin/activate
PYTHONPATH="$PWD/python:$PWD/python/lib/py4j-0.10.9.9-src.zip" python -m unittest \
  pyspark.tests.test_install_spark.SparkInstallationTestCase.test_checked_versions \
  pyspark.tests.test_install_spark.SparkInstallationTestCase.test_package_name

Additional checks passed:

• Ruff lint and format checks on both changed files.
• Standalone mypy checking of the full install.py source using python/mypy.ini, without importing the parent PySpark package.
• git diff --check.
• Search confirmed no remaining UNSUPPORTED_COMBINATIONS references under python.

The network-dependent installation/download test was not run. An earlier package-aware mypy invocation encountered 23 errors elsewhere in PySpark; this is not a claim that the full package type check passes.

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

Generated-by: pi 1.1.0

[gaogaotiantian]

pyspark.install should be internal only - I did not find it in our documentation. It's okay if we remove a global variable from it if it's not considered public. The variable has not been used for 6 years and we should just remove it.