[Enhancement] Add Semantic Kernel plugin wrapper for Deepgram functions

[deepgram-robot]

Summary

Add a Semantic Kernel plugin class that exposes Deepgram STT, TTS, and Audio Intelligence as [KernelFunction] methods, enabling .NET AI agents built with Microsoft Semantic Kernel to natively invoke Deepgram capabilities.

Problem it solves

Microsoft Semantic Kernel (53k+ stars) is the standard framework for building AI agents in .NET. Developers creating voice-enabled agents, meeting assistants, and audio processing copilots need Deepgram functions registered as kernel plugins so the AI can autonomously decide when to transcribe audio, generate speech, or analyze sentiment. Without this, .NET agent developers must write manual glue code between Semantic Kernel and Deepgram.

Proposed API

using Deepgram.SemanticKernel;

var kernel = Kernel.CreateBuilder()
    .AddDeepgramPlugin(apiKey: "...")
    .Build();

// Plugin exposes:
// [KernelFunction] TranscribeAudioAsync(string audioUrl) -> string
// [KernelFunction] SpeakTextAsync(string text, string voice) -> byte[]
// [KernelFunction] AnalyzeSentimentAsync(string text) -> SentimentResult

Acceptance criteria

• Plugin class with [KernelFunction] decorated methods for STT, TTS, and Intelligence
• Function descriptions are AI-friendly (agents can discover and select appropriate functions)
• Supports both URL-based and file-based audio input for transcription
• Documented with usage example showing an agent chat loop
• Compatible with existing Deepgram .NET SDK API

---

Raised by the DX intelligence system.