The memory used to hold shuffle descriptors must be unused during the
entire execution of the shuffle sequence, and therefore needs to be
free at input *and* output time.

Once we have picked a region for the shuffle descriptors, we can free
this working copy and start over with a clean memory map, the only
thing we do there is mark as reserved the memory used by the
descriptors.

Only use 64 as a static guarantee of the number of shuffle
descriptors; corresponding to 768 bytes of memory.

Query the maximum number of shuffle descriptors, and use that number
as the descriptor block size.

It is safe to assume that the number of moves will never decrease as
the reserved memory space increases; thus, no need to do individual
increments; skip ahead.

Allocate high memory out of the way to hold the shuffle descriptors,
and generate continuation descriptors as needed.

All the internal symbols in mempcpy.S were "memcpy", which is
obviously wrong.

Update copyright lines and standardize the format.

Conflicts:

	extlinux/extlinux.c

size == 0 means delete item; we don't want a header then.

size > 255 is always an error.