The printing code would check the same environment byte multiple times and
write to the console one byte at a time.  For some devices (such as the
Blackfin JTAG console which operates in 8 bytes at a time), this is pretty
damned slow.  So create a small 16 byte buffer to fill up and send to puts
as needed.  In the process, unify the different print functions, shrink
the resulting code (source and compiled), and avoid excess env reads as
those too can be somewhat expensive depending on the board.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>