Relocate spark table logic from the block processor to a dedicated directive scanner. This allows spark tables to be treated as `:md-table` directives and ensures they are processed in a second pass after the content index is fully populated.
