Fix #11: local variable 'time_points' referenced before assignment

This fixes the issue #11 (closed) which were two related bugs:

• There was an assumption that if frequency_hz is present, then monotonic_time_points_us can not be present.
• There was a second assumption that if calendar_time_points is present, then monotonic_time_points_us must be present.

The way I understand the spec, both assumptions are invalid.