feat: forward generation params via MITM + add usageMetadata to Gemini

- Add GenerationParams struct to MitmStore for temperature, top_p,
  top_k, max_output_tokens, stop_sequences, frequency/presence_penalty
- MITM modify_request injects params into request.generationConfig
- All 3 endpoints (Completions, Responses, Gemini) store client params
- Add usageMetadata to Gemini sync responses (promptTokenCount,
  candidatesTokenCount, totalTokenCount, thoughtsTokenCount)
- Add generation param fields to GeminiRequest (temperature, topP, etc.)
- Completions stream_options.include_usage emits final usage chunk
- Completions reasoning_tokens in completion_tokens_details
- Update endpoint gap analysis doc (all high-priority gaps resolved)

src/mitm/store.rs

@@ -60,6 +60,21 @@ pub struct PendingToolResult {
     pub result: serde_json::Value,
 }
 
+/// Client-specified generation parameters for MITM injection.
+/// Set by API handlers, consumed by the MITM modify layer.
+#[derive(Debug, Clone, Default)]
+pub struct GenerationParams {
+    pub temperature: Option<f64>,
+    pub top_p: Option<f64>,
+    pub top_k: Option<u32>,
+    pub max_output_tokens: Option<u64>,
+    pub stop_sequences: Option<Vec<String>>,
+    /// Frequency penalty (OpenAI) — mapped to frequencyPenalty in Gemini.
+    pub frequency_penalty: Option<f64>,
+    /// Presence penalty (OpenAI) — mapped to presencePenalty in Gemini.
+    pub presence_penalty: Option<f64>,
+}
+
 /// Thread-safe store for intercepted data.
 ///
 /// Keyed by a unique request ID that we can correlate with cascade operations.
@@ -102,6 +117,10 @@ pub struct MitmStore {
     captured_thinking_text: Arc<RwLock<Option<String>>>,
     /// Whether the captured response is complete (finishReason received).
     response_complete: Arc<AtomicBool>,
+
+    // ── Generation parameters for MITM injection ─────────────────────────
+    /// Client-specified sampling parameters to inject into Google API requests.
+    generation_params: Arc<RwLock<Option<GenerationParams>>>,
 }
 
 /// Aggregate statistics across all intercepted traffic.
@@ -144,6 +163,7 @@ impl MitmStore {
             captured_response_text: Arc::new(RwLock::new(None)),
             captured_thinking_text: Arc::new(RwLock::new(None)),
             response_complete: Arc::new(AtomicBool::new(false)),
+            generation_params: Arc::new(RwLock::new(None)),
         }
     }
 
@@ -433,4 +453,21 @@ impl MitmStore {
     pub async fn clear_active_cascade(&self) {
         *self.active_cascade_id.write().await = None;
     }
+
+    // ── Generation parameters ────────────────────────────────────────────
+
+    /// Store client-specified generation parameters for MITM injection.
+    pub async fn set_generation_params(&self, params: GenerationParams) {
+        *self.generation_params.write().await = Some(params);
+    }
+
+    /// Read current generation parameters (non-consuming).
+    pub async fn get_generation_params(&self) -> Option<GenerationParams> {
+        self.generation_params.read().await.clone()
+    }
+
+    /// Clear generation parameters.
+    pub async fn clear_generation_params(&self) {
+        *self.generation_params.write().await = None;
+    }
 }