After TIG-2656 is done, ensure the DSI and non-DSI versions of the microbenchmarks workloads show comparable performance numbers for at least 2 waterfall runs. If they do, remove the non-DSI tasks and remove the _dsi suffix from the DSI tasks.

Get buy-in from the performance team and/or the perf-build-barons via slack prior to merging.

At the same time, make perf.yml and sys-perf.yml look as identical as possible. The only difference at this point should be the list of tasks and the names of the variants. Everything else should be the same now.