The multitask generator sometimes fails to produce points when a suggest can be for more points than are in the buffer. To work reliably with libEnsemble, this should be able to generate more points to fill out the request.
This may include generating extra sample points.
Without this test, it could results in inefficient scenarios where not all workers are kept busy, or in hangs, if no points are produced.
A gest-api generator should error if cannot return the number of points requested.