Write terraform for long-running testing

[ethan-readyset]

Description

To ease the process of running long-running tests in the future, we want to write some terraform to spin up the necessary resources and run a sysbench workload against ReadySet.

We should be able to spin up the following resources:

• Serverless RDS Aurora Postgres instance with a configurable scaling threshold
• EC2 instance of configurable size in which we run ReadySet
• EC2 instance of configurable size in which we run sysbench
• A cloudwatch dashboard that charts the following:
  • ReadySet EC2 Instance: CPU, memory, disk I/O, network throughput, ReadySet-specific memory metrics
  • RDS instance: write/read IOPS, write throughput/read throughput, oldest replication slot size, ACU utilization, volume write IOPS

Additionally, all the logs and metrics from the instances/RDS instance should be sent to cloudwatch and should be available persistently even after the resources are torn down after the test. We should also be able to run an initial iteration of the test without ReadySet to establish a baseline.

Change in user-visible behavior

Requires documentation change

[ethan-readyset]

I made some progress on this but ultimately ended up just using the pre-existing infrastructure we already had set up from the previous long-running tests. I'm backlogging this for now since there is more pressing work