Add Anthropic prompt caching (direct + via OpenRouter)

Caches the system prompt/tools and growing conversation history via
cache_control breakpoints, cutting cost and latency on repeated turns.
Covers both the regular chat path and the tool-calling loop
(chatWithToolMessages), which has its own request-building code and was
initially missed. Cost calculation now accounts for cache write/read
pricing instead of treating all input tokens as full price. Verified
live: cache reads grow turn-over-turn in oAI.log.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

oAI/Providers/OpenRouterModels.swift

@@ -48,7 +48,12 @@ struct OpenRouterChatRequest: Codable {
     let toolChoice: String?
     let modalities: [String]?
     let reasoning: ReasoningAPIConfig?
-    
+    let cacheControl: CacheControl?
+
+    struct CacheControl: Codable {
+        let type: String
+    }
+
     struct APIMessage: Codable {
         let role: String
         let content: MessageContent
@@ -138,6 +143,7 @@ struct OpenRouterChatRequest: Codable {
         case toolChoice = "tool_choice"
         case modalities
         case reasoning
+        case cacheControl = "cache_control"
     }
 }
 
@@ -225,11 +231,23 @@ struct OpenRouterChatResponse: Codable {
         let promptTokens: Int
         let completionTokens: Int
         let totalTokens: Int
-        
+        let promptTokensDetails: PromptTokensDetails?
+
+        struct PromptTokensDetails: Codable {
+            let cachedTokens: Int?
+            let cacheWriteTokens: Int?
+
+            enum CodingKeys: String, CodingKey {
+                case cachedTokens = "cached_tokens"
+                case cacheWriteTokens = "cache_write_tokens"
+            }
+        }
+
         enum CodingKeys: String, CodingKey {
             case promptTokens = "prompt_tokens"
             case completionTokens = "completion_tokens"
             case totalTokens = "total_tokens"
+            case promptTokensDetails = "prompt_tokens_details"
         }
     }
 }