Make the tests less relaxed

The behavior of jinja2 seems to be different between linux and osx
4 jobs for dict-sort in 8 minutes and 23 seconds (queued for 4 seconds)
Status Job ID Name Coverage
  Build
passed #126984
docker
build_linux_27

00:07:25

54.0%
passed #126985
docker
build_linux_36

00:06:44

53.0%
passed #126986
macosx
build_macosx_27

00:08:23

54.0%
passed #126987
macosx
build_macosx_36

00:07:33

53.0%