RMLTC0019b/RMLTC0020b: Test cases silently assume a best-effort operation

[mcndt]

There are two ways an RML Mapper can handle data errors in a source:

1. Try to ignore those records and continue mapping (best effort)
2. Fail when a record cannot be mapped to a IRI-safe triple (strict)

For cases such as RMLTC0019b and RMLTC0020b, the test assumes the first mode of operation (discovered through the MySQL and PostgreSQL variants of the test case).

This is different from the R2RML variant of the test case, which assumes the second mode of operation.

Perhaps these assumptions should be documented?