Consider transitioning our front-end fleet from ssd to standard disks

The transition from standard disks to SSD happened with the GCP migration. One of the main benefits of this was not performance, but that there was a dramatic decrease in deployment time due to the large ~1GB omnibus package installation.

With the move to K8s maybe we should we re-evaluate the need to use SSD. For the front-end there we shouldn't be writing too much to disk (where we might do more for sidekiq due to temporary files).

Perhaps we should switch a cluster to using pd-standard instead of pd-ssd to see if there is an impact? We will want to see how much it will have an effect on pod start times, for example.

This might also have a cost benefit.

Related to production#4940 (closed) where we believe there was no more SSD to allocate in us-east1-b.

Dropping this in teamDelivery but we may want to move it to teamReliability .