Add parsing for detailed token usage information (#282)

This update adds functionality to parse the detailed token usage data returned by the OpenAI API.
Reference: https://platform.openai.com/docs/api-reference/chat/object#chat/object-usage

Author: e-max

async-openai/src/types/chat.rs

@@ -93,6 +93,32 @@ pub struct CompletionUsage {
     pub completion_tokens: u32,
     /// Total number of tokens used in the request (prompt + completion).
     pub total_tokens: u32,
+    /// Breakdown of tokens used in the prompt.
+    pub prompt_tokens_details: PromptTokenDetails,
+    /// Breakdown of tokens used in a completion.
+    pub completion_tokens_details: CompletionTokenDetails,
+}
+
+/// Breakdown of tokens used in a completion.
+#[derive(Debug, Deserialize, Serialize, Clone, PartialEq)]
+pub struct PromptTokenDetails {
+    /// Audio input tokens present in the prompt.
+    #[serde(default)]
+    pub audio_tokens: u32,
+    /// Cached tokens present in the prompt.
+    #[serde(default)]
+    pub cached_tokens: u32,
+}
+
+/// Breakdown of tokens used in a completion.
+#[derive(Debug, Deserialize, Serialize, Clone, PartialEq)]
+pub struct CompletionTokenDetails {
+    /// Audio input tokens generated by the model.
+    #[serde(default)]
+    audio_tokens: u32,
+    /// Tokens generated by the model for reasoning.
+    #[serde(default)]
+    reasoning_tokens: u32,
 }
 
 #[derive(Debug, Serialize, Deserialize, Default, Clone, Builder, PartialEq)]