Rework how warmup kernels are specified and run

[rhornung67]

Summary

• This PR makes it easier to specify and run warmup kernels in different ways (turn-off warmups, our default feature-based warmup strategy, specify a specific set of kernels to run for warmup, or warmup all kernels that will be run when the suite runs).
• Note that this PR introduces a simple, but admittedly unseemly change involving a static variable in the KernelBase class to make warmup kernels run only once regardless of the specified # run reps or kernel default run reps. We can change this later to something more desirable with more involved code changes.
• Also, fixed some output file table header names.

[rhornung67]

@MrBurmark I'm going to spice this PR up a bit by adding some more options to the -wk command line arg to make it easier to work with.