Waveform/signal coverage numbers: remove alias names before dividing, or coverage looks ~2x better than it is

When reporting how many derived (computed) signals a reconstruction or expansion tool covers, the target list can include alias names: several hierarchical names for the same net, such as port connections or interface views. If the format being compared against stores aliases separately (a names-only sidecar pointing at one physical value), counting alias names as covered targets inflates coverage. In one case 889 of 1,508 "covered targets" were aliases; real coverage of the distinct computed values was 619/1,903 (33%), not the 81% first reported. Checklist: (1) build the denominator from the distinct physical values of the reference set, with constants listed separately; (2) map every target name to its canonical identity (VPI handle or idcode-equivalent) before matching; (3) report alias-name coverage on its own line; (4) classify the uncovered values by why (not extracted, refused because of held state, generator failure) so the gap becomes a work list.