cls/rbd, librbd: use uint64_t for sparse_size

cls/rbd, librbd: use uint64_t for sparse_size

the size of `size_t` is varies from architecture to architecture. the
C++ standard only requires it to be able to represent the maximum possible
size of object of any type. on 32-bit architectures, it's very likely a
32-bit unsigned integer. to ensure the interoperability between the
64-bit systems and 32-bit systems, we should use a type with explicitly
defined size.

also, we don't define the dencoder for size_t. so on systems where
size_t is not backed by uint32_t or uint64_t, the tree does not compile.

in this change, use uint64_t for sparse_size. and leave
`C_SparsifyObject::m_sparse_size` intact. as the latter should be able
to be promoted to uint64_t when necessary.

this change is backward compatible on 64-bit systems.

Signed-off-by: Kefu Chai <kchai@redhat.com>