Benchmark pipeline improvements

[juliasloan25]

Is your feature request related to a problem? Please describe.
The benchmark pipeline is helpful, but flaky. Namely, runs frequently hang and then time out, and the timing numbers we get vary a lot from run to run. We have some ideas below for ways to reduce the number of (non-meaningful) failures in this pipeline.

To do

• decrease pipeline slurm time for the runs to 15 mins (bucket, richardsmodel) or 20 mins (soil/canopy, snowy land) to reduce wait time if runs hang
• make hanging runs soft fail (exit code -1)
• increase the accepted bounds for the timing value from +/- 2 to +/- 5 standard deviations

[charleskawczynski]

Where are the benchmark files located?