The VkAccelerationStructureInfo structure is serialised followed by the unified geometry readback buffer.

When the initial state Apply() is called the ASes are built from the input data one at a time so that a single scratch buffer can be built (enlarging when needed) and re-used.  Although this is slower it is necessary on Mali as it has poor space efficiency for the scratch buffer and so can easily OOM a device if all the ASes are built in a single command buffer submission.

Rather than have a buffer per geometry, allocate a single block to carry all the input geometry.

Change-Id: Ibc6f221ba0a317e8006ccf4a252da71c0aefe7dc

drbrdt

Change-Id: Ia41dd97fec4f7a2b78fc505dacc0304bbc818789

If an initial content AS was destroyed during active capture then resetting before serialisation meant that VkInitialContents::Free() was called, deref-ing the AS info and causing it to be deleted.

So VkInitialContents is now reset after serialisation to ensure its lifetime.

Change-Id: I6a296bd96dd5b639f68347073298e5545e8f980d

Then clone to captured AS.  This is much faster than building on each Apply().

Also:
* Use GPUBuffer instead of own radback memory mechanism, but mark as internal
* Use GPU local buffer for AS builds instead of upload memory to improve performance