feat(ai): improve runtime capability reporting and context settings

Add explicit chat capability tracking, expose formatted runtime capabilities
in the info command, and support a max context size option for user AI settings.

Also update Ollama base URL resolution to use OLLAMA_ADDRESS and simplify
provider chat command execution.

src/ai/ai-runtime-target.ts

@@ -74,7 +74,7 @@ function modelEnvNames(provider: AiProvider, purpose: AiRuntimePurpose): string[
 export function getProviderBaseUrl(provider: AiProvider): string | undefined {
     switch (provider) {
         case AiProvider.OLLAMA:
-            return env("OLLAMA_ENDPOINT");
+            return env("OLLAMA_ADDRESS");
         case AiProvider.GEMINI:
             return env("GEMINI_BASE_URL") ?? env("GEMINI_ENDPOINT")
                 ?? (Environment.GEMINI_API_MODE === "openai" ? GEMINI_OPENAI_BASE_URL : undefined);