Persistent embedding cache

Write a Python function embed_with_cache(texts, embed_model, namespace='', cache_path=None) that returns the embeddings for the given texts using the cached_text_embeddings function from the kg.memory_common module. The function should ensure that repeated calls with the same texts and same model produce cache hits (no new embedding computation). In the function's docstring or a comment, explain why the float32 round-trip (array.array('f', vec).tobytes() then frombytes) is performed inside cached_text_embeddings to maintain precision symmetry between cache hits and misses. You may assume embed_model has a get_text_embedding_batch method. Provide a brief demonstration of the cache behavior (e.g., call the function twice with the same texts and note that only the first call triggers embedding). Use the real identifiers and module paths from the provided repository source.