Separate verification from performance tests

[Nanoseb]

Right now, tests both check the performance by computing BW, timing, running the same test n_iter times, as well as the accuracy of the solution by comparing the results to a known analytical solution.

Even if the implementation of the test is similar, the two type of tests require different input parameters. i.e. no need to run the same computation 1000 times to check its accuracy and performance tests require larger meshes for example.

Several options could be used:

• Writing distinct tests depending on the goal and have a cmake option to switch between one set or the other.
• For each feature, having both tests in a single file and change the inputs based on:
  • an environment variable
  • a command line parameter
  • a compile time variable
  • ...?
• other option?

Any other option to consider? @pbartholomew08 @semi-h

[Nanoseb]

(I think I'd vote for an environment variable as that's the easiest to handle and use in this kind of scenario)