Add selectable TTS engine (Windows/Piper/Kokoro) with Kokoro neural voices

[willk]

I felt that piper voices were a little harsh. I added Kokoro. I did use a robot to help since I don't have any experience in .net.

I've tried a bunch of other tools for EQL and nothing has been as good at alerting as EQLogParse.

Add selectable TTS engine (Windows / Piper / Kokoro), with working Kokoro neural voices

Changes

EQLogParser.Audio (audio engine)

• AudioManager: replaced the implicit "first engine found wins" startup logic with an explicit, persisted engine preference (WindowsEngine / PiperEngine / KokoroEngine constants). Initialize() now takes a preferredEngine parameter; if unset it falls back to the previous auto-detect order (Piper → Kokoro → Windows) for existing installs.
• Added AudioManager.GetAvailableEngines() — lists engines that can actually be selected right now (Windows always; Piper if its voice pack is present; Kokoro if the model has been downloaded).
• Added AudioManager.GetActiveEngine() — reports which engine is actually running this session, so the UI can flag when a change needs a restart.
• PiperTts: added IsVoicePackAvailable(), a side-effect-free check for the voice pack (previously the only way to check was calling Initialize(), which has side effects like setting the native DLL search path).
• IAudioManager: added GetActiveEngine() to the public contract.

EQLogParser (app)

• App.xaml.cs: reads the persisted "TtsEngine" setting and passes it into AudioManager.Initialize() at startup.
• New TtsEngineWindow (replaces KokoroSetupWindow): adds a combo box to pick the active TTS engine from whatever's currently available, persists the choice, and shows a "restart required" hint when the selection differs from the active engine — alongside the existing Kokoro model download flow.
• TriggersView: renamed the "Kokoro TTS" button to "TTS Engine" and pointed it at the new window.

Installer

• EQLogParserInstall.iss: minor — installs the bundled Kokoro voices\* folder unconditionally (previously only Piper's voices were referenced).

Behavior after this change

• Fresh/existing installs keep today's behavior by default (Piper wins if present).
• Downloading Kokoro and explicitly selecting it in the new TTS Engine window makes it the active engine on next launch, regardless of whether Piper is also present.
• Once Kokoro is active, the existing voice dropdowns (Triggers tab default voice, per-character voice in Manage Characters, WAV export) list its ~50 voices — that plumbing already existed and just needed Kokoro to actually get a turn.

Not covered

• No automated tests added (none existed for AudioManager's engine selection previously).
• Not manually verified end-to-end in a packaged installer build — tested via local dev build only.

[kauffman12]

Yeah this is a good idea. I wanted to add something better than piper at some point. I know it's not amazing but at least it's small and would run on anything. But I don't think anyone really likes it and having a few options would be good. I'll try this out.

[kauffman12]

So looking at this I may want to have piper work the same way and then make a lot of the voices optional to cut down on how much needs to get installed. Plus I'm not sure if the fp32 model might be overkill. For simple speech I don't know if anyone could tell the difference if it uses fp16 or something instead. Still a good idea just make take a little bit to integrate everything.