openai: Don't send prompt_cache_key for OpenAI-compatible models (#36231)

Some APIs fail when they get this parameter

Closes #36215

Release Notes:

- Fixed OpenAI-compatible providers that don't support prompt caching
and/or reasoning

crates/language_models/src/provider/vercel.rs

@@ -355,6 +355,7 @@ impl LanguageModel for VercelLanguageModel {
             request,
             self.model.id(),
             self.model.supports_parallel_tool_calls(),
+            self.model.supports_prompt_cache_key(),
             self.max_output_tokens(),
             None,
         );