Load BL31 to DDR instead of On-Chip RAM for scalability. Also, make use
of On-Chip RAM for BL31 specific variables filling down from handoff
offset to reduce fragmentation

Signed-off-by: Hadi Asyrafi <muhammad.hadi.asyrafi.abdul.halim@intel.com>
Change-Id: Ib64f48bd14f71e5fca2d406f4ede3386f2881099