ai: add some new env variables to control OpenAI requests

[GRbit]

Here are some changes/updates to code that handles OpenAI requests.

• default model changed to gpt-4o-mini cause it's cheaper and better than gpt-3.5-turbo (see models description and pricing)
• new OPENAI_API_MODEL env variable
• new OPENAI_API_NUMBER_COMPLETIONS variable
• new OPENAI_API_SYSTEM_PROMPT variable
• &http.Client{} in shared/ai replaced with http.DefaultClient. Go http clients are perfectly ready for using from multiple goroutines, there is no need in allocated a new client each time you do http request.

Don't hesitate to change anything, request any changes, or share some suggestions.