improve heat autoscaling test

[Pavlo Shchelokovskyy]

current heat autoscaling test in OSTF has couple of shortcomings:
- CPU load and release commands are (AWKward) bash one-liners passed via SSH, very much dependent on image/shell/SSH implementation, prone to bugs
- the command being used is only good for loading a single vCPU core

I propose to stabilize it somewhat by placing an only 'sh' dependent script on instances during boot and launching it via SSH with appropriate switches, my own simplest and Cirros/TestVM tested examples here
https://github.com/pshchelo/stackdev/blob/284d51900153674fd6aaa143e8f7cb5a38394856/scripts/cpuload
https://github.com/pshchelo/stackdev/blob/c950e1ffce6198b926941fa48de04702fdd7815f/templates/autoscaling/asg.yaml

Another small advantage is that such test stack template also exercises SoftwareConfig resource and get_file template function to some extent.

Test can also be improved to use WaitConditions and nested stacks.

[Alexander Nagovitsyn (gluk12189)]

now we use other metrics for scale instance. The test has been rewritten
No changes are required.