improve upgrade part of role re drain & restart

[tarzanek]

Sometimes drain can get "stuck"
and retries to check for it will fail - we should have a way to override this check

then again stopping of Scylla process will also get blocked - so this waits for systemd unnecessarily long (15mins or so), we should have a way to kill it immediatelly or after certain delay (given as input)

above can be caused by various bugs - so role should be flexible to give you and override
in a sense above harmless since commitlog is saving the day - which then in turn means, if you do above, you have to wait for commitlog replay
(so start of node will take longer, which is OK)

we need to get above flexibility for easier upgrades from versions with drain bugs