CLOUDP-280378 is going to enable us to run e2e performance tests using gRPC by modifying the Envoy config to skip the gRPC => MongoRPC transcoder in Atlas. However, we likely do not want to push those configuration options to mms master, as we would never want this enabled in Atlas.

To accomplish this, we want to run mms, with the changes to skip the transcoder, locally, and then spawn a cluster from mms with a running mongot process in order to successfully run the performance tests. This should not make the performance numbers any flakier than if we ran them in evergreen, as either way the performance workload is is running in the cloud. 

Using the following test (query-insurance-25GB-text-and-exists.yml#L186-L212) on a M30-replset-AWS-S20_2 cluster type (with the latest mongod binary compiled for ARM), generate the initial performance numbers for MongoRPC in evergreen, and confirm with edwin.jacox@mongodb.com that these numbers are comparable to the "run mms locally and spawn the cluster from local mms" approach. This will give us a baseline set of performance numbers to compare gRPC to, and validate our performance testing strategy for this project.