fix(spark): match Spark 4 encode charset behavior - #27260

Summary

Addresses apache/datafusion-spark#46 by adding selectable Spark 4-compatible behavior to the datafusion-spark encode expression.

• Adds datafusion.spark.legacy_java_charsets and datafusion.spark.legacy_coding_error_action, both defaulting to false.
• Enforces Spark 4's seven canonical charset names by default while retaining legacy aliases when legacy_java_charsets is enabled.
• Raises MALFORMED_CHARACTER_CODING (SQLSTATE 22000) for unmappable characters by default.
• Restores legacy replacement behavior when legacy_coding_error_action is enabled.
• Adds unit tests and SQLLogicTests for charset handling, configuration combinations, error behavior, and settings exposure.
• Documents both configuration options.

Validation

• cargo check -p datafusion-spark -p datafusion-common
• cargo clippy -p datafusion-spark -p datafusion-common -- -D warnings
• cargo fmt --check
• Focused encode unit tests
• Full datafusion-spark library tests (333 passed)
• Focused encode.slt SQLLogicTest
• Focused information_schema.slt SQLLogicTest
• git diff HEAD^ HEAD --check

Compatibility note

The existing empty-input UTF-16 behavior is unchanged: DataFusion emits the UTF-16 BOM, whereas Spark returns an empty byte array. This pre-existing divergence is documented and retained.

Closes apache/datafusion-spark#46.