feat: switchable router embedder and L3 backends via config

Adds [router.embedder] and [router.l3] sections to CorrexConfig with
backend selectors. Ships LlamaCppEmbedder that hits llama.cpp's
/embedding endpoint (handles OpenAI-compatible, simple, and array
response shapes; validates dimension). InfrastructureModule gains
createEmbedderFromConfig and createL3MemoryStoreFromConfig that
dispatch on backend value.

Defaults preserve current behavior (noop embedder + in-memory L3).
Switching to "llamacpp" / "turbovec" is a config-only change — no code
edits required. For turbovec backend, the bundled python sidecar
script is extracted from classpath to ~/.cache/correx/ on first use.

apps/server/src/main/kotlin/com/correx/apps/server/Main.kt

@@ -124,12 +124,16 @@ fun main() {
         sandboxRoot = sandboxRoot,
         defaultSystemPromptPath = toolsConfig.defaultSystemPromptPath,
     )
-    val routerConfig = RouterConfig()
+    val routerConfig = correxConfig.router
+    val embedder = InfrastructureModule.createEmbedderFromConfig(routerConfig.embedder)
+    val l3MemoryStore = InfrastructureModule.createL3MemoryStoreFromConfig(routerConfig.l3)
     val routerFacade = InfrastructureModule.createRouterFacade(
         eventStore = eventStore,
         inferenceRouter = inferenceRouter,
-        config = routerConfig,
+        config = RouterConfig(),
         tokenizer = llamaProvider.tokenizer,
+        embedder = embedder,
+        l3MemoryStore = l3MemoryStore,
     )
     val module = ServerModule(
         orchestrator = orchestrator,