Ollama post-processing silently skipped in service/toggle mode

[tindotdev]

Describe the bug
When using whis toggle (service mode), Ollama post-processing is silently skipped even when post_processing.enabled: true and processor: "ollama" are configured. The same configuration works correctly in CLI mode (whis --post-process).

To reproduce

1. Configure Ollama as post-processor with processor: "ollama" and enabled: true
2. Configure services.ollama.url and services.ollama.model
3. Run whis toggle and use hotkey to transcribe
4. Transcription returns without post-processing (no error shown)

Expected behavior
Ollama post-processing should work in service mode, same as CLI mode.

Environment

• OS: Fedora 43 (Cosmic desktop, Wayland with wl-copy)
• Whis version: 0.7.2
• Installation method: cargo install

Additional context
Root cause: In service.rs:422-449, post-processing requires api_key() to return Some, but api_key() returns None for Ollama (since it uses a URL, not an API key). This causes the post-processing branch to be skipped.

Suggested fix: Handle Ollama separately using URL instead of checking for API key.

Workaround: Use OpenAI or Mistral as post-processor instead.