Voice Plugin: Support custom/remote provider (URL) instead of local-only processing

[SLi95c]

Problem

LeanType already supports connecting external AI providers (Gemini, Groq, OpenAI-compatible, and local GGUF models) for text processing like proofreading and translation. The Voice Plugin, however, is currently limited to local/on-device processing only, with no option to configure a custom endpoint or URL.

Requested Feature

Add the ability to configure a custom provider/URL for voice processing, similar to the existing "OpenAI-compatible" approach for text — for example:

• Custom server URL (e.g. self-hosted Whisper, faster-whisper, or an Ollama-based speech model)
• API key field for cloud services (e.g. OpenAI Whisper API, Groq Whisper)
• A toggle between "Local" (current behavior) and "Remote/Custom Endpoint"

Motivation

• Speed: Audio processing on more powerful server/desktop hardware is typically much faster than on-device processing on a phone.
• Quality: Larger/better speech models (e.g. Whisper large vs. a small mobile-optimized model) produce noticeably more accurate transcriptions.
• Consistency with existing features: Text-based AI already supports external/self-hosted providers, so a similar option for voice would make the app's feature set more consistent.
• Hardware independence: Users with older or lower-end devices would benefit from offloading processing, without losing the feature entirely.

Proposal

Add to the Voice Plugin settings:

1. A field for a custom endpoint URL
2. An optional API key field
3. A clear privacy disclaimer that audio is sent to an external service when a remote provider is used — matching the existing disclaimer for text AI providers

Local processing would remain the default, so the app's privacy/offline model stays intact — remote processing would be strictly opt-in.