Store official GH pipeline benchmarks separately

[jordandsullivan]

Within the run_benchmarks.sh script, we just save everything in the ucc/benchmarks/results folder. It would be nice if we had a way to separate these official runs (on the GH action) from local testing runs we might do, or indeed the previous local runs from before we implemented the pipeline.

One way to do this is to modify the benchmark script something like

#!/bin/bash

if [ "$GITHUB_ACTIONS" == "true" ]; then
    # Define the results folder path
    RESULTS_FOLDER="$SCRIPT_DIR/../results/pipeline"
else
    # Define the results folder path
    RESULTS_FOLDER="$SCRIPT_DIR/../results"
fi

This would also solve the problem of file conflicts when writing to datafiles created on the same day (initially this is something I did to not end up with a bunch of individual datafiles when running multiple compiler x benchmark combinations in parallel.

In #133 I change the ucc/benchmarks/scripts/save function to only append to the same file if they were run in the same hour on the same day, which seems like a reasonable compromise to me.

[bachase]

An alternative would be to have two styles of automated runs

1. Per release, do the existing style run and store the results back to the repo. I believe Update benchmark plot resolution to monthly #200 will be doing this.
2. Add a separate action for any PR, that runs the benchmarks and adds a comment back to that PR showing the delta from the current baseline run. (Something like codecov does)This results would never be store in the repo, but github would have enough info to tell us what (if any changed) from that specific PR. This would also avoid needing to cleanup/delete commits from any manual runs. We might have to add some special tags, or limited which users can trigger a run, to avoid abuse of the action.

For any local testing, we could either not have it save the benchmark results at all, or save to a separate directory that .gitignore prevents from sharing.