Configurable context size for completions

[saulotoledo]

Currently, ECA sends a large amount of context for ghost completions, which can slow down the response time, especially when using local models like Ollama. To improve performance, it would be helpful to have a dedicated setting to control the number of lines or context size sent specifically for ghost completions, separate from other interactions (e.g., chat or explicit requests).

[ericdallo]

Thanks for the suggestion @saulotoledo!
The thing is that ECA already send for completion a small prompt and only that here so not sure what you think it could be improved there