Identify process required when replacing perf machines in the infrastructure

Since performance results are dependent on the machines that are used for performance, we should put a process together to document what we should do when we are required to migrate to alternate (usually newer) hardware in order to avoid any misinterpretation of performance results.

Recently we [replaced a number of the s390x machines](https://ci.adoptopenjdk.net/label/ci.role.perf&&hw.arch.s390x/) so they may have already shown some performance differences, but we are also in the process of replacing the machines at Equinix which host our [ci.role.perf&&hw.arch.x64&&sw.os.linux systems](https://ci.adoptopenjdk.net/label/ci.role.perf&&hw.arch.x86&&sw.os.linux/) which will see similar changes (Although both of those are currently offline - did we remove the need for perf jobs to run with that tag?)

No source files or tests are named. Start by reviewing the linked s390x and x64 Linux performance-machine labels and the current perf jobs that use them, then document the migration steps needed to preserve interpretation of results and clarify whether those jobs still need the referenced tag.

Written by the indexing model from the issue text.