[GH-3281] GeoPandas: implement from_features

[jiayuasu]

Did you read the Contributor Guide?

• Yes, I have read the Contributor Rules and Contributor Development Guide.

Is this PR related to a ticket?

• Yes, and the PR name follows the format [GH-XXX] my subject.

Closes #3281.
Part of #2230.

What changes were proposed in this PR?

This PR implements GeoDataFrame.from_features for small, in-memory GeoJSON-like feature collections.

It:

• delegates feature parsing to the installed GeoPandas version and wraps the resulting frame in Sedona;
• accepts feature iterables and generators, FeatureCollection dictionaries, and objects implementing __geo_interface__;
• preserves column selection and ordering, CRS assignment, null geometries, uneven or null property sets, and empty inputs with explicit columns;
• preserves leading-null geometry order on Spark 3, where pandas-on-Spark otherwise fails Geometry UDT inference; and
• documents driver materialization, Spark property-schema constraints, and the distributed GeoJSON and GeoParquet alternatives for large datasets.

Malformed inputs continue to use GeoPandas validation and error behavior. Property values must be representable by a consistent Spark-inferred schema after parsing.

How was this patch tested?

• Spark 3.4.0 / Python 3.8 / GeoPandas 0.13.2: 3 focused tests passed.
• Spark 3.4.0 / Python 3.8 / Shapely 1.8.5.post1: 3 focused tests passed.
• Spark 4.1.1 / Python 3.11 / GeoPandas 1.1.3: 3 focused tests passed.
• All commit hooks passed.

The tests cover supported input forms, generators, __geo_interface__, leading-null geometries, uneven and null properties, column ordering, CRS preservation, restored row order, and empty inputs with an explicit geometry schema.

Did this PR include necessary documentation updates?

• Yes, I am adding a new API. I am using the current SNAPSHOT version number in vX.Y.Z format.

The method documentation marks the API as added in 2.0.0, and the release notes are updated.