add $options['missing']='none' and 'adjusted' to similarity_classical 'memory' operations

Computation in-memory is much faster than in-database. But recommender_similary_classical() only supports in-memory computation when missing data is treated as zero. In-memory computation is needed for cases where missing data is treated as 'none' or 'adjusted' like the in-database operation.

This is a low priority code task because the full functionality is already implemented in in-database approach. The in-memory feature is just an optional feature rather than a necessary one.