Autocomplete: add deepseek-coder:6.7b support

[valerybugakov]

Context

• Adds autocomplete support for local inference with deepseek-coder:6.7b-base-q4_K_M powered by ollama.
• Adds a bit of abstraction to simplify further addition of new ollama models.
• Sets deepseek-coder:6.7 as a default model for the ollama provider.
• Should be tested with the follow-up perf improvements Autocomplete: disable parallel inference requests for ollama #2967

Test plan

1. Install and run Ollama
2. Download one of the support local models:
   • ollama pull deepseek-coder:6.7b-base-q4_K_M for deepseek-coder
   • ollama pull codellama:7b-code for codellama
3. Update Cody's VS Code settings to use the unstable-ollama autocomplete provider.
4. Confirm Cody uses Ollama by looking at the Cody output channel or the autocomplete trace view (in the command palette).