test against a self-hosted Sentry in GCP

- benchmarking
- what fails in Sentry (or GCP) first if you send too many events. Would GCP even let that many requests reach your Sentry instance?
- quantify how many events and over what timerange(s) this happened
- Billing, set cut-offs
- https://github.com/getsentry/onpremise
- AM data in addition to EM