fpga: zynq: Correct fpga load when buf is not aligned

When ARCH_DMA_MINALIGN is greater than header size of the bit file, and buf is
not aligned, new_buf address became greater then buf_start address and the
load_word loop corrupts bit file data.

A work around is to decrease new_buf of ARCH_DMA_MINALIGN, it might corrupt data
before buf but permits to load correctly.

Signed-off-by: Stany MARCEL <smarcel@novasys-ingenierie.com>
Signed-off-by: Michal Simek <michal.simek@xilinx.com>