Currently the scroll copies the full screen which is slow on high
resolution displays. At the same time, most of the screen is an empty
space which has no need to be copied over and over.

Optimize the scrolling algorithm by caching the x coordinates of the
last printed lines and scroll in accordance with the maximum x in that
cache.

On my Microsoft Surface Book (the first version) this produces a
significant speedup of the console 90 seconds vs. 168 seconds with the
kernel command line having

	ignore_loglevel earlycon=efifb keep_bootcon
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>