Correctly size prometheus disks on all clusters

[yuvipanda]

Until #3093 was merged, we had a massive explosion in the size of prometheus disk usage. We had kept increasing the sizes to keep up. Now that disk size has stabilized, we should evaluate them and size them back down.

Generally resizing disks down is not possible, but given #3119 will involve recreating most disks on GCP, we should use this opportunity to size them back down. Given that pd-balanced is more expensive than pd-standard, sizing them down while doing this migration should help us keep costs more stable