Octavia deployment failed because of missing qemu-img

Octavia deployment started failing with an error related to image management:

./logs/undercloud/home/zuul/standalone-ansible-mlkg0y7x/octavia-ansible/octavia-ansible.log:2020-12-02 03:44:10,188 p=150211 u=root n=ansible | 2020-12-02 03:44:10.187212 | bc764e10-346e-b786-3ba0-0000000000a7 | FATAL | convert image from qcow2 to raw | standalone | error={"changed": true, "cmd": "qemu-img convert -f qcow2 -O raw /home/zuul/amphora.qcow2 /tmp/ansible.23wp1trx/amphora.img\n", "delta": "0:00:00.003573", "end": "2020-12-02 03:44:10.158330", "msg": "non-zero return code", "rc": 127, "start": "2020-12-02 03:44:10.154757", "stderr": "/bin/sh: qemu-img: command not found", "stderr_lines": ["/bin/sh: qemu-img: command not found"], "stdout": "", "stdout_lines": []}

It looks like qemu-img was previously installed during the install_package.sh stage (as a ceph-ansible dependency?). It now needs to be installed during Octavia deployment steps.