fix: exclude "off" from reasoning_effort in requesty provider

The OpenAI API types do not include "off" as a valid reasoning_effort value,
so we need to exclude it similar to how we handle "minimal"

Author: roomote

src/api/providers/requesty.ts

@@ -123,7 +123,9 @@ export class RequestyHandler extends BaseProvider implements SingleCompletionHan
 			model,
 			max_tokens,
 			temperature,
-			...(reasoning_effort && reasoning_effort !== "minimal" && { reasoning_effort }),
+			...(reasoning_effort &&
+				reasoning_effort !== "minimal" &&
+				reasoning_effort !== "off" && { reasoning_effort }),
 			...(thinking && { thinking }),
 			stream: true,
 			stream_options: { include_usage: true },