mkinitfs: add threaded zstd compression

-T0 defaults to hardware cores (no SMT cores, so 8 on a 8core/16thread
cpu)

for something small like the initrd, so many threads can't be utilised
anyway, so generally this will only use a few (3-4). it does make it
take much less time, however

$ zstd -b19 out
19#out : 86131024 -> 22705519 (x3.793), 6.05 MB/s, 1395.7 MB/s
$ zstd -T0 -b19 out
19#out : 86131024 -> 22636531 (x3.805), 15.4 MB/s, 1394.2 MB/s