spi flash: fix two off-by-one bugs that were cancelling out

c_status_time was one cycle longer than it should be!
this led to the arria5 requiring g_dummy_time-1 in c_vstatus_data
==> after fixing: must increase g_input_to_output_cycles on arria5

there was also a wrong calculation for s_wip with single-lane mode
it was working also by luck because WEL and WIP are usually set
together and are side-by-side, thus masking the c_status_time bug.