Add realloc support

[sim642]

Closes #701.

Also includes a couple of blob and pointer casting changes that were necessary to get chrony's array.c to work soundly (i.e. not going to VD.top) while not breaking any other tests (particularly with calloc).

TODO

• Tasks from Complete realloc support #701.
• Figure out what to do about cast_addr changing alloc variable type from void. Had to disable to get chrony array.c to work (otherwise a cast to char* screwed up the type and any following casts to a larger struct type were considered not ok, so supertop got introduced).
• Adapt realloc accesses to free type after Add option for ignoring races from free #695 merge.

[sim642]

The small cast_addr change of not changing varinfo types exposes a profound amount of trouble. So many things seem to be based on types (which we don't have for allocated memory) instead of on the structure of the abstract value. These mismatching for alloc variables (and possibly other cases) cause countless operations to just fall all the way back to their default supertop-returning case and transitively spreading those supertops everywhere else as well.

[sim642]

I got all tests to pass again by hacking the varinfo type-change hack even further: avoid it when casting to char* because that's a special pointer cast by the standard and used for byte-based access to memory (like chrony's array.c is doing).