Waveform capture: alias signals dominate callback count; serve them from their source primary

When profiling which signal classes drive per-change callback cost in a simulator capture (VPI/DPI value-change callbacks), split the extra signals beyond the primaries (nets actually driven by logic) into aliases (another name for the same net, e.g. ports and hierarchical clock connections) and combinational outputs. In one design, aliases were ~63% of the extra calls, and the top entries were all clock aliases, each firing every cycle. Adding the aliases on top of the primaries cost ~32% more total CPU; adding the combinational signals cost ~28% (marginal, not additive). An alias always equals its source, so if the source is already recorded, the alias can be resolved at read time by an index remap, with no subscription and no compute. Do that before optimising the collector itself: packing/sorting tweaks saved only ~3% total CPU because they don't reduce the number of callbacks.