[SPARK-56188][PS] Align Series.map({}) with pandas 3 empty-dict behavior

[ueshin]

What changes were proposed in this pull request?

This PR updates pandas-on-Spark Series.map for the pandas 3 case when the mapper is an empty plain dict.

In pandas 3, pandas.Series.map({}) returns an all-NaN float64 Series. pandas-on-Spark was preserving the input Spark string type for this path, so a string Series produced a pandas StringDtype result instead. This PR adds a pandas-3-only branch in pyspark.pandas.Series.map so Series.map({}) returns a null DoubleType column for an empty plain dict, which materializes as the same all-NaN float64 result as pandas.

The related test was updated to check the pandas 3 expectation directly, while keeping the pre-pandas-3 expectation unchanged. The existing defaultdict coverage remains in place so dict subclasses with __missing__ continue to follow the existing behavior. This patch also includes a small test adjustment so the local helper used in test_map only uppercases string inputs.

Why are the changes needed?

Without this change, pandas-on-Spark diverges from pandas 3 for Series.map({}). For example, a string Series currently returns a pandas StringDtype result in pandas-on-Spark, while pandas returns an all-NaN float64 Series.

Aligning this narrow case avoids a pandas-version-specific mismatch in pandas-on-Spark behavior and keeps the test expectation consistent with pandas 3 internals.

Does this PR introduce any user-facing change?

Yes, it will behave more like pandas 3.

How was this patch tested?

Updated the related tests.

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

Generated-by: Codex (GPT-5)

[dongjoon-hyun]

+1, LGTM.

[dongjoon-hyun]

Merged to master for Apache Spark 4.2.0.