Allow passing non heap objects to WeakRef

[jhass]

This should allow at least pointers to the data section to work seamlessly with WeakRef and in some cases even allow stack pointers to be passed, for the crazy. This should increase the usability of WeakRef since it allows passing static data for testing or other purposes.

This also renames WeakRef#target to WeakRef#get WeakRef#value. get seems to be the more general, more widely used and shorter name. value is consistent with Pointer. Additionally target feels like it's leaking implementation details as
that's the name of the internal instance variable too.

As expected this is slightly slower for the heap case, given the additional check, and slightly faster for the non-heap case, given the check is less expensive than registering a pointer with GC. A quick benchmark with collection disabled so the second testcase wouldn't crash:

            WeakRef heap  14.42M ( 69.36ns) (±11.41%)       fastest
        WeakRef constant  10.55M (  94.8ns) (±45.52%)  1.37× slower
    CheckingWeakRef heap   9.14M ( 109.4ns) (±45.29%)  1.58× slower
CheckingWeakRef constant  11.48M ( 87.07ns) (±83.27%)  1.26× slower

[bcardiff]

Java use #get. C# use #target.

I think I picked from C# back then.

I could agree that changing to #value since it would match the Pointer(T) method. But I feel that #get is too generic.

[jhass]

I could live with value, what do others think?

[RX14]

#value to be consistent with Pointer looks good to me.

[asterite]

I like #value too :-)

[asterite]

@jhass Nice! I always test WeakRef with WeakRef.new("foo") and see it crash and then remember "oh, it's not in the heap". It's something ugly that should probably not crash at all, so this PR is very welcome! :-)