fix for #2 -- CodeLlama crashes

- add replacement tokenizer class for unknown tokenizers
- fix quantization for models that don't have lm_head quantized

Requires https://github.com/ml-explore/mlx-swift/pull/28

Libraries/LLM/Llama.swift

@@ -187,11 +187,13 @@ public class LlamaModelInner: Module {
 
 public class LlamaModel: Module, LLMModel {
 
+    public let vocabularySize: Int
     let model: LlamaModelInner
 
     @ModuleInfo(key: "lm_head") var lmHead: Linear
 
     public init(_ args: LlamaConfiguration) {
+        self.vocabularySize = args.vocabularySize
         self.model = LlamaModelInner(args)
         self._lmHead.wrappedValue = Linear(args.hiddenSize, args.vocabularySize, bias: false)
     }