perf: evaluate new machine types in GCE, AWS, Azure

### Background

Currently, we run nightlies with the following machine types, chosen according to a specified (random) distribution,

| machine_type | cloud |
|-------------------|-------|
| n2 | GCE |
| n2d | GCE |
| t2a | GCE |
| m7g | AWS |
| c7g | AWS |
| c6a | AWS |
| c6i | AWS |
| m6i | AWS |
| r6i | AWS |
| Standard_Dxds_v5 | Azure |
| Standard_Dxlds_v5 | Azure |
| Standard_Dxpds_v5 | Azure |
| Standard_Exds_v5 | Azure |

Note, the above machine types were extracted from `Select(GCE|AWS|Azure)MachineType` and the corresponding data-driven test [1]. It's been nearly two years since the above machine type instances were evaluated [2]. Since then, each cloud has deployed new machine types.

### Motivation

The new machine types may potentially yield better cost per performance on some of the workloads. The high-level goal is to determine what those workloads might be as well as the potential trade-offs. As a first approximation, we should plan to evaluate the following machine types,

| machine_type | cloud |
|-------------------|-------|
| n4, n4d | GCE |
| n4a, c4a | GCE |
| c4 | GCE |
| m8g | AWS |
| c8g | AWS |
| c8a | AWS |
| c8i | AWS |
| m8i | AWS |
| r8i | AWS |
| Standard_Dxds_v7 | Azure |
| Standard_Dxlds_v7 | Azure |
| Standard_Dxpds_v7 | Azure |
| Standard_Exds_v7 | Azure |
| Standard_Dplds_v6 (arm) | Azure |

[1] https://github.com/cockroachdb/cockroach/blob/master/pkg/cmd/roachtest/testdata/cluster_test/machine_types
[2] https://github.com/cockroachdb/cockroach/pull/117852

Jira issue: CRDB-56388

Read the Select(GCE|AWS|Azure)MachineType definitions and the data-driven test in pkg/cmd/roachtest/testdata/cluster_test/machine_types. Compare the existing machine-type list with the proposed GCE, AWS, and Azure instances, then review PR 117852 for prior evaluation context. Done means evaluating the listed machines for workload performance and cost trade-offs.

Written by the indexing model from the issue text.