Fair waveform-dump overhead comparisons: make every format dump the same signal set, and check it from the outputs

When comparing a simulator's native waveform dump against a DPI-based capture into another format, the easiest unfairness is a different signal set: e.g. native $dumpvars(0, scope) dumps a whole hierarchy while the DPI arm captures a handful of selected signals. Make the native arm list the same signals explicitly ($dumpvars/$fsdbDumpvars with each signal path), not a scope, and check it from the outputs: decode every file and compare the signal count and event count against one reference reader, so an arm that dumped more or fewer signals fails the check. Keep whole-scope dumping as a separate scaling point, because there the native path (built into the simulator) and the DPI path (a call per changed group) have different per-change costs, and that is exactly what you want to measure. Also include a no-dump arm and a capture-only or null-sink arm, so you can separate tap or DPI cost from format and compression cost.