Beginning with Python 3.5, an lru_cache no longer survives a deepcopy.

```
$ cat >> cache_copy.py
import copy
from functools import lru_cache
foo = lru_cache()(lambda: None)
copy.deepcopy(foo)
$ python3.5 cache_copy.py
Traceback (most recent call last):
  File "cache_copy.py", line 4, in <module>
    copy.deepcopy(foo)
  File "/usr/lib/python3.5/copy.py", line 182, in deepcopy
    y = _reconstruct(x, rv, 1, memo)
  File "/usr/lib/python3.5/copy.py", line 293, in _reconstruct
    y = callable(*args)
  File "/usr/lib/python3.5/copyreg.py", line 88, in __newobj__
    return cls.__new__(cls, *args)
TypeError: Required argument 'user_function' (pos 1) not found
```

I suspect this is due to the C implementation per issue14373.

I realize one's first reaction might be "why are you deep copying a function," so here's the [downstream bug](https://bitbucket.org/jaraco/jaraco.functools/issues/1/method_cache-causes-typeerror-when) and [offended code](https://bitbucket.org/jaraco/jaraco.functools/src/68d1fda21af7e7b4c36577f953f382270bdc1e05/jaraco/functools.py?at=default&fileviewer=file-view-default#functools.py-72:138). That decorator is designed to wrap a method and to store the cache on the instance object, an object which is liable to be deep copied. As a result, production code using this technique is failing on Python 3.5.

Although it was apparently an artifact of previous implementations that the cache happened to be deep-copyable, I hope this expectation can be restored with minimal fuss.