add ability to define deployment strategy for unicorn and sidekiq

In case when unicorn or sidekiq deployments do not have time to finish pods recreating gitlab operator starts the whole circle anew: shared-creds job, migrations, deployments update because the operator has deployments waiting timeout 300 seconds.
Related issue #1941 (closed)
I propose to add ability to define unicorn and sidekiq deployments strategy to be able to make the rolling update process faster.
I suppose it will be useful not only in case with the operators timeout.