Lazy slot fix reintroduces storage leak

[gbracha]

The fix preserve s the method metadata during serialization. However the metdata we needed to preserve is a LazySlotTag, which is a nested class that drags in all kinds of stuff with it. This defeats the fix we recently introduced for deployment. So we'll need to clean this up.

[gbracha]

Changing metadata to a more general class that is part of the kernel, along with some adjustments to the serializer, solves the leak for vfuels. For nso files, lazy slots still cause a leak. THis probably requires a bit more fiddling with serialization, or else a cruder metadata representation (like an array).