Defensively use crossJoin syntax rather than setting crossJoin.enabled to true

As a fix for issue #892 we configure `spark.sql.crossJoin.enabled` to `true` in both Mapping Rule interpreters.

This configuration then remains true for the remainder of the application life cycle.

This is quite a dangerous thing to do as it may hide future bugs and warnings, which would otherwise raise AnalysisException in tests etc.

Much better approach would be to:
- investigate this bug a little further
- it's currently assumed, that this issue only arises when there's a single join condition using a column previously created by the literal rule
- it's only reproducible locally when the data and mapping tables are read from files
- we should analyse the list of conformance rules and in cases, where there's a mapping rule with a single join condition (one of which was previously created by literal) we should use the crossJoin syntax explicitly rather than left_outer join