Deleting space makes `body.space` raise `ReferenceError`

[aatle]

Garbage collecting the space a body was added to causes the body's _space proxy to die. But, the Body.space property does not catch or check whether the proxy is alive, unlike the Shape.space property.
Additionally, the Body.sleep/sleep_with_group methods might not work as intended when the space is deleted.

Example:

import pymunk as pm

space = pm.Space()
body = pm.Body()

space.add(body)

del space

print(body.space)

Traceback (most recent call last):
  File "C:\Users\...\main.py", line 10, in <module>
    print(body.space)
          ^^^^^^^^^^
  File "C:\Users\...\pymunk\body.py", line 425, in space
    return self._space._get_self()  # ugly hack because of weakref
           ^^^^^^^^^^^^^^^^^^^^^
ReferenceError: weakly-referenced object no longer exists

Suggestion:
I think it is better to use weakref.ref instead of weakref.proxy.
When there is no space to reference, use a None returning callable (types.NoneType, lambda: None, weakref(set())). (In my own package, I use the latter that is cached as a global.) Then, having a deleted space is treated the same as not having a space in the first place.
Conveniently, the return type is already Optional[Space].

-       if self._space is not None:
-           try:
-               return self._space._get_self()  # ugly hack because of weakref
-           except ReferenceError:
-               return None
-       else:
-           return None
+       return self._space()

(Internal code might now use .space instead of ._space so it doesn't need to invoke the weakref itself.)

[aatle]

I forgot to mention, the above suggestion was also towards Shape._space, as it'll simplify the code, be consistent, and remove the need for Space._get_self().

[viblo]

Ah of course. I thought about it but then forgot it before I had time to implement. Now its fixed (but again forgot the issue tag..)