Slab size computation needs to be changed for atree inlining

[fxamacker]

Problem

This isn't a problem yet, but it will surface with atree inlining.

When the same value is reset in Array or OrderedMap and its storable is a pointer, slab size computation can be wrong by simply taking a size diff of existing element and new element. This is because existing and new elements are the same and they returns the same storable size, which results in 0 size diff, even if element content is modified and actual size is different.

Suggestion

Change the way slab size is computed in Set functions to avoid the scenario described.

[fxamacker]

This is closed by PRs #336 and #337.