Do not initialize ReadBufferFromS3 for queries needing only virtual columns

Why it matters

This feature addresses the inefficiency of initializing data reading buffers from S3 storage when queries request only virtual columns, which do not require actual data loading from S3. It improves query performance and resource usage by avoiding unnecessary network and I/O operations.

How to use it

This optimization is applied automatically when executing queries that access only virtual columns from S3 data sources, requiring no additional user configuration.