More C# alignment

[jsm174]

Follow-ups from reviewing the recent AI-submitted PRs, aligning the port with the C# source and fixing the issues those reviews surfaced:

• pacled64: resolve the unit via map::at(), matching the C# dictionary indexer (throws for a missing key, caught in Pinball::MainThreadDoIt exactly like C#'s KeyNotFoundException).
• misc (AlarmHandler): the no-parameter RegisterAlarm now unregisters same-handler alarms before adding, per AlarmHandler.cs (the parameterized overload already did). Renames the parameter to doNotUnregister (C# DoNotUnregister) and drops the stored flag C# doesn't keep. Closes the stale MaxRunTimeMotorStop accumulation possible after Fix Motor toy never scheduling its kickstart/max-runtime alarms #55.
• misc (CabinetOutputList): GetByName's controller+number fast path now falls through to the by-name loop when the numbered output doesn't exist, matching the C# indexer; TryParseInt now requires full consumption like int.TryParse ("1x" no longer parses as 1).
• misc (LedWizEquivalent): remove the comment and now-unused includes left by Resolve LedWizEquivalent outputs via CabinetOutputList (fixes PinOne + FTDI "0 of N resolved") #60.
• pinone: connect failure messages and log level now match PinOne.cs exactly.
• pico: m_oldOutputValues initialized to 255 per C# (Enumerable.Repeat((byte)255, ...)), so the initial all-off sweep actually transmits.
• misc (auto-configurators): restore the C# hierarchical {ctrl}\{ctrl}.NN output name form in LedWiz, PacLed64, PacUIO, Pinscape, PinscapePico, and DudesCab configurators (safe now that resolution goes through GetByName after Resolve LedWizEquivalent outputs via CabinetOutputList (fixes PinOne + FTDI "0 of N resolved") #60); align FT245R's format string with the C# original. PacDrive stays flat, matching C#.
• pacuio: replace the unit vector with a map keyed like the C# static dictionary (0-2), fixing out-of-bounds UB at Init/Update/OnOutputValueChanged for a <PacUIO> with a missing Id; outputs are now added via OutputList::Add so they subscribe to value-changed events.
• pinone (NamedPipeServer): supersedes Fix PinOne DOF output silently dropped on Linux (sp_open return ignored) #53. sp_open's result was ignored, so a bad or unopenable port left the server running with a dead handle, silently dropping every write (10.8.1 Linux standalone: DOF not working with PinOne Control Board vpinball#2962). The constructor now throws, making CreateServer return false and ConnectToController fail the same way C# does when SerialPort.Open() throws. The server thread previously blocked forever in accept(), hanging StopServer()'s join() on every teardown ("Updater thread did not quit" in 10.8.1 Linux standalone: DOF not working with PinOne Control Board #31); select() with a 200ms timeout bounds both accept() and the client read so the thread re-checks m_isRunning — unlike Fix PinOne DOF output silently dropped on Linux (sp_open return ignored) #53's SO_RCVTIMEO, this also works on macOS, where that option does not affect accept() (verified empirically). Port close/free is now owned solely by StopServer, removing the pre-existing double sp_close.

Testing: full build clean; teardown behavior of the select-based server loop verified on macOS with a standalone harness (no-client teardown joins in ~110ms where accept() previously blocked forever; silent-client teardown ~105ms; data path unaffected).

[dynajoe]

Thanks for bringing this all together. I tried to keep each fix isolated and small to make incremental progress.

I've been jamming on replacing my Windows cabinet with a linux version this past week and it's looking very promising.

[jsm174]

Thanks for bringing this all together. I tried to keep each fix isolated and small to make incremental progress.

I've been jamming on replacing my Windows cabinet with a linux version this past week and it's looking very promising.

What hardware do you have? We have a discord channel that would be helpful for feedback.

[dynajoe]

Thanks for bringing this all together. I tried to keep each fix isolated and small to make incremental progress.
I've been jamming on replacing my Windows cabinet with a linux version this past week and it's looking very promising.

What hardware do you have? We have a discord channel that would be helpful for feedback.

I'm running:

• OS: Batocera 43
• Frontend: vpinfe
• Graphics: RTX 3080
• Audio: SSF
• Playfield: LG 43" 4K OLED 120hz
• separate 1080p displays for backglass and dmd
• Solenoids / buttons powered by Cleveland Software PinOne
• Headtracking: Kinect V2