Arrow Flight data source support

[Neutrinic]

Extend the connector to support Arrow Flight endpoints.

Requirements:

• Connect to Arrow Flight servers using FlightClient
• Schema discovery via FlightInfo/getSchema
• Stream FlightData directly into VectorSchemaRoot (skip JSON→Arrow conversion)
• Support Flight SQL for JDBC-like sources exposed via Flight

Design considerations:

• Most efficient path: Arrow Flight already speaks Arrow, so no conversion needed
• Reuse Spark integration layer (RESTScan, PartitionReader) but bypass JSON/HTTP layers
• Config: type = flight, endpoint = "grpc://host:port", command = "..."
• Could support ColumnarBatch directly for zero-copy (unlike REST which uses InternalRow)

Blocked by: Manual REST config workflow complete