tests: correct the output order about starting a background thread for Windows
I didn't track down where this change occurred. I assume it's related to some
buffering changes, and/or an explicit flush somewhere.
Differential Revision: https://phab.mercurial-scm.org/D9539
$ hg init
$ touch unknown
$ touch a
$ hg add a
$ hg ci -m "1"
$ touch b
$ hg add b
$ hg ci -m "2"
Should show unknown
$ hg status
? unknown
$ hg revert -r 0 --all
removing b
Should show unknown and b removed
$ hg status
R b
? unknown
Should show a and unknown
$ ls -A
.hg
a
unknown