Cast away volatile on dereference in CAS-based generalization primitives

Cast away volatile on dereference in CAS-based generalization primitives

* src/atomic_ops/generalize-arithm.template
(AO_XSIZE_fetch_and_add_full, AO_XSIZE_fetch_and_add_acquire,
AO_XSIZE_fetch_and_add_release, AO_XSIZE_fetch_and_add,
AO_XSIZE_and_full, AO_XSIZE_or_full, AO_XSIZE_xor_full): Cast away the
volatile (when reading old value before CAS) for architectures like
IA64 (where volatile adds barrier semantics).
* src/atomic_ops/generalize-small.template (AO_XSIZE_load_read,
AO_XSIZE_load_full, AO_XSIZE_load_acquire, AO_XSIZE_load): Likewise.
* src/atomic_ops/generalize-arithm.h: Regenerate.
* src/atomic_ops/generalize-small.h: Likewise.