Connectivity: Prevent duplicate logical pin assignments

[Andrei-Errapart]

Summary

Two connectivity defects let a component's inactive or duplicated graphics distort the extracted netlist. Pins belonging to a display mode other than the component's selected one still participated in connectivity, so an alternate symbol graphic (notably dual-row header alternates) could capture adjacent nets. And a symbol exposing the same physical package pad at multiple graphical locations produced multiple independent logical pins, so one pad's endpoints could land in different nets or a net could list the same pin more than once.

Changes

• SheetGraph now skips pins whose OwnerPartDisplayMode differs from the component's selected DisplayMode, extending the existing inactive-part (CurrentPartId) filter to inactive display modes.
• New SheetGraph.UnifyDuplicatePhysicalPins unions all graphical endpoints that share a physical package pad (component + pin designator), so they form one electrical node even when drawn at different locations.
• NetlistAssembler emits one logical pin per sheet instance and pad, deduplicating by SheetInstanceId plus pin key when assembling each net's pin list.
• New tests in SchematicNetlistTests covering both behaviors.

Test Plan

• Existing tests pass (dotnet test) — full suite at this commit: 853 passed, 0 failed, 10 skipped (pre-existing data-dependent skips, unrelated to this change)
• New tests added for new behavior — Only_Selected_Display_Mode_Pins_Participate_In_Connectivity (Theory, run for both display modes: each mode's pin connects only to its own net and the netlist contains the logical pin exactly once) and Duplicate_Pin_Designators_On_Component_Unify_By_Physical_Pad (both graphical locations of pad 6 resolve to a single net listing the pin once)

Checklist

• Code follows existing style and conventions
• Public API changes are documented with XML doc comments — there are no public API changes: NetlistAssembler and SheetGraph are internal, the new UnifyDuplicatePhysicalPins is private, and the properties the tests use (SchPin.OwnerPartDisplayMode, SchComponent.DisplayMode) already existed
• No breaking changes (or clearly identified below) — no API signature changes. Netlist extraction output does change for affected schematics, which is the fix itself: pins of unselected display modes no longer join nets, and endpoints sharing a physical pad now merge into one logical pin instead of appearing as duplicates.

[Andrei-Errapart]

CI for this branch passed on my fork (same workflow, same base commit): https://github.com/Andrei-Errapart/AltiumSharp/actions/runs/32387331575