image.py: don't call layer_tar.getmembers() twice in _squash_layers

[vrutkovs]

This should save a couple of seconds on large builds

[goldmann]

Just a side note from testing, your code performs like this with an example image:

[root@performance docker-scripts]# time PYTHONPATH=. python docker_scripts/cli.py squash -f 7427c9af1454 -t image:squashed image:latest 2> /dev/null

real    0m21.112s
user    0m6.653s
sys 0m4.190s
[root@performance docker-scripts]# time PYTHONPATH=. python docker_scripts/cli.py squash -f 7427c9af1454 -t image:squashed image:latest 2> /dev/null

real    0m21.198s
user    0m6.796s
sys 0m4.292s
[root@performance docker-scripts]# time PYTHONPATH=. python docker_scripts/cli.py squash -f 7427c9af1454 -t image:squashed image:latest 2> /dev/null

real    0m20.901s
user    0m6.520s
sys 0m4.250s

Master is at:

[root@performance docker-scripts]# time PYTHONPATH=. python docker_scripts/cli.py squash -f 7427c9af1454 -t image:squashed image:latest 2> /dev/null

real    0m21.347s
user    0m6.561s
sys 0m4.264s
[root@performance docker-scripts]# time PYTHONPATH=. python docker_scripts/cli.py squash -f 7427c9af1454 -t image:squashed image:latest 2> /dev/null

real    0m21.505s
user    0m6.705s
sys 0m4.155s
[root@performance docker-scripts]# time PYTHONPATH=. python docker_scripts/cli.py squash -f 7427c9af1454 -t image:squashed image:latest 2> /dev/null

real    0m19.983s
user    0m6.770s
sys 0m4.239s

Not that big difference :)

[goldmann]

Thanks, merging!

Related to #63.