Build a Decorator That Memoizes Pure Functions

Memoization (caching of function results) is a powerful technique to avoid recomputing expensive pure function calls when inputs repeat. Build a Python decorator named memoize that: - Caches results per-call based on the function's positional and keyword arguments. - Uses a tuple of args and a deterministic representation of kwargs as the cache key (e.g., tuple(sorted(kwargs.items()))). - Preserves the wrapped function's metadata using functools.wraps. - If any part of the key is unhashable (for example, lists inside args), the decorator should not raise an error; instead it should call the original function without caching and return its result. Your decorator must work for ordinary functions (including recursive ones like a memoized fibonacci). Keep the implementation simple and self-contained; thread-safety is not required for this exercise.

Write a decorator @memoize that caches call results for functions called with hashable positional and keyword arguments, while safely handling unhashable arguments by bypassing the cache.

Cache only when a deterministic, hashable key can be built from (args, kwargs). Do not attempt to serialize arbitrary objects. If construction of the key raises TypeError (unhashable component), call the original function directly without caching. Use functools.wraps to preserve metadata.