Add possibility for malloc to fail (via setting)

[schuessf]

Previously, our C translation assumed that malloc (and the other allocation functions, calloc and realloc) could never fail and return NULL. For SV-COMP, it was decided that this assumption should be dropped (see !1686 for the corresponding benchmark changes and discussion). Therefore, this PR includes the following changes:

• A setting Assume that allocation on the heap always succeeds was added. When disabled (the new default), allocation may return NULL in our translation, otherwise the behavior remains unchanged.
• The translation was adapted to reflect this setting: For malloc, the procedure contract was changed (see NonDetStrategy2D and SimpleIncreasingStrategy), while for calloc and realloc the body in the translation was changed (see StdlibLibraryModel and ConstructRealloc).
• There are two places where we generate malloc calls internally, for "non-standard" purposes, that still rely on the allocation never failing: varargs handling (see FunctionHandler) and non-deterministic string generation (see FunctionModelHelper). Rather than introduce a separate allocation function, I took the simpler route of just adding an explicit assume statement after these two calls.
• Since the new setting defaults to false, several regression tests needed explicit NULL-handling so they don't fail after this change.