Introduce skip_upgrade_check option

A funny situation arose when we introduced the option to fail early when Scylla is already installed and upgrade_version == False:

It may happen that the role failed during its runtime, therefore leaving the target systems in an inconsistent state. When this happens, if the failure happened AFTER the Scylla package was installed, then the user would be left in an chicken-and-egg situation:
They can not re-execute the playbook installing the specified version, as the upgrade will abort since we are upgrading to the same release. Similarly, they can not install Scylla again, as it will complain that the package is already installed.

We circumvent that situation by introducing the `skip_upgrade_check` option, which is a failsafe mechanism to allow a role failure to proceed on installation even when `upgrade_version` is False. Needless to say, this option is False by default.

ansible-scylla-node/defaults/main.yml

@@ -5,6 +5,12 @@
 # UPGRADE
 # ==============================
 
+# Skips role upgrade check. Useful if the role failed in the middle of an installation.
+# Values:
+# [default] false: Fail when Scylla is installed and the upgrade_version == False
+# true: When upgrade_version == False, re-run the install process 
+skip_upgrade_check: false
+
 # Upgrades Scylla version if a previous installation is detected
 # Values:
 #  [default] false: don't check for upgrades

ansible-scylla-node/tasks/main.yml

@@ -13,8 +13,9 @@
     msg: >
       Detected an existing Scylla installation.
       Should you want to upgrade, set `upgrade_version' to True.
+      If you are recovering from a failed role run, set `skip_upgrade_check' to True.
   when: >
-      not upgrade_version and
+      not upgrade_version and not skip_upgrade_check and
       ( 'scylla-server' in ansible_facts.packages or
         'scylla-enterprise-server' in ansible_facts.packages )