[m-rev.] for review: add builtin 8, 16 and 32 bit integer types -- Part 2

[Zoltan Somogyi]

After doing a "git pull" of this change, I started getting warnings like this
on my Linux laptop (which has gcc 4.8.1):

erlang_rtti_implementation.c:6504:23: warning: cast from pointer to integer of different size [-Wpointer-to-int-cast]
           Int16_37 = ((int16_t) conv8_Int16_37);


I get the warnings both when the .c file is generated by the compiler
from before this change, and when it is generated by the compiler
from after this change.

The warning is for all casts from values of type MR_Box, which is defined
as "void *" for the MLDS backend, to any signed or unsigned integer type
smaller than a pointer.

Julien, can you please add an intermediate cast to MR_Integer to such code?
I expect you have the right place to do so paged in right now.

Zoltan.