role fails in check mode if not having been executed in normal mode before

[pcahyna]

Followup to #27 .

We fixed failures in check mode some time ago and now the role does not fail if executed in check mode after having been executed in normal mode. There is still a problem though when the role is executed in check mode on a fresh system (without having been applied in normal mode before), because it assumes that some packages are installed, and in check mode they do not get installed. This is the case described in ansible/molecule#128 (comment) (note that the current Molecule setup does not test for it, as explained in that link).
The change submitted by @hhaniel in #27 (comment) should fix this case as well.
Test playbook:

- name: Ensure that the role runs with default parameters
  hosts: all
  roles:
    - name: timesync
      vars:
        timesync_ntp_provider: ntp

error:

TASK [timesync : Get ntp version] **********************************************
fatal: [/home/pcahyna/linux-system-roles-testing/local/CentOS-7-x86_64-GenericCloud-1702.qcow2c]: FAILED! => {"changed": false, "cmd": ["rpm", "-q", "--qf", "%{version}", "ntp"], "delta": "0:00:00.033978", "end": "2019-09-30 14:53:22.430978", "msg": "non-zero return code", "rc": 1, "start": "2019-09-30 14:53:22.397000", "stderr": "", "stderr_lines": [], "stdout": "package ntp is not installed", "stdout_lines": ["package ntp is not installed"]}

PLAY RECAP *********************************************************************
/home/pcahyna/linux-system-roles-testing/local/CentOS-7-x86_64-GenericCloud-1702.qcow2c : ok=3 changed=1 unreachable=0 failed=1 skipped=9 rescued=0 ignored=0

[pcahyna]

Note that the test uses the ntp provider, because chrony is installed in that image, therefore the problem does not show up with the default provider, at least not with the CentOS image I used.