fcompare: Add option for an absolute tolerance check

[sayerhs]

This PR adds a command line option to fcompare that takes in a user-specified tolerance for absolute error and adds logic to compare both absolute and relative errors against user-specified tolerances.

Since -a is already used as a short-form for --allow_diff_grids, I have only used --abs_tol as the option without a short version when parsing command line options.

Additional background

For ExaWind/AMR-Wind we would like to compare our nightly regression tests executing on GPUs against plot files generated on CPUs. However, due to indeterministic execution of nodal projection on GPUs the solutions are slightly different for each GPU run. In most cases, the relative tolerance specification is sufficient. However, for particular cases when the absolute error is really small, there are situations when relative error exceeds the specified tolerance and the check fails, as shown in the following fcompare output. Here gpx, gpy, velocityy, velocityz are failing checks. Having an absolute tolerance check will allow skipping these corner cases.

--------------------------------------------------------------------------

            variable name            absolute error            relative error
                                        (||A - B||)         (||A - B||/||A||)
 ----------------------------------------------------------------------------
 level = 0
 density                                          0                         0
 gpx                                1.108446668e-14           1.669216956e-10
 gpy                                 5.96855898e-15           1.675179285e-07
 gpz                                9.048317651e-15           4.122654211e-14
 mu_turb                            7.160938509e-15           2.134236096e-14
 p                                  1.136868377e-13           3.862064706e-15
 temperature                        4.547473509e-13           1.697082323e-15
 tke                                4.468647674e-15           4.855703851e-14
 velocityx                          1.065814104e-14           1.332264646e-15
 velocityy                          3.714614144e-15           1.064078537e-10
 velocityz                           1.17857113e-14           1.076369457e-10
 velocity_mueff                     7.160938509e-15           2.134236096e-14

Checklist

The proposed changes:

• fix a bug or incorrect behavior in AMReX
• add new capabilities to AMReX
• changes answers in the test suite to more than roundoff level
• are likely to significantly affect the results of downstream AMReX users
• are described in the proposed changes to the AMReX documentation, if appropriate