Fix result_scan columns for empty records batch

[osipovartem]

• Added schema serialization to ResultSet to handle empty record batches

  • Previously, when the ResultSet contained no rows (empty record batches), the schema was inferred at runtime using infer_json_schema, which sometimes resulted in an empty schema. This caused issues downstream because a valid schema is required even when no data rows are present (e.g., for joins).
  • To address this, extended the ResultSet struct to include a serialized schema field. This allows the schema to be preserved and reused directly, ensuring consistent and complete schema information regardless of the presence of rows.

• Additionally, enabled the serde feature for arrow-schema to support serialization and deserialization of Schema. This is necessary because DataFusion itself does not enable this feature by default, so we had to explicitly activate it in our build to make schema (de)serialization possible.

Closes #1058 #1037
Related to #1057

[github-actions]

SQL Logic Tests Results ❌

Coverage by SLT File

• to_json: 0/4 (0.0%)
• parse_json: 2/8 (25.0%)
• result_scan: 0/3 (0.0%)

Overall: 2/15 (13.3%)