Avoid Reusing Scalar Subquery Cache in MV Block Processing

Avoid reusing the scalar subquery cache when processing MV blocks. This fixes a bug when the scalar query reference the source table but it means that all subscalar queries in the MV definition will be calculated for each block. #32811 (Raúl Marín).