Fix for the second-order PLM slope reconstruction.

[magomedovs]

This MR fixes the second-order PLM slope reconstruction (MC-limited centered slope), which is selected by setting pelec.plm_iorder = 2. Previously, selecting this option resulted in the following relationship for the dtemp variable: dtemp = 4.0 / 3.0 * dcen rather than dtemp = dcen. This led to the overestimation of smooth slopes and the production of an artificial gradient steepening.
To reproduce the issue, run the "PeleC/Exec/RegTests/EB-C12" regression test case with pelec.do_mol = 0 and pelec.plm_iorder = 2.

[d-montgomery]

@magomedovs , thank you for opening up this PR and bringing this to our attention. I'm testing the changes now and reviewing the literature. I'll let you know if I have any questions.

[magomedovs]

@magomedovs , thank you for opening up this PR and bringing this to our attention. I'm testing the changes now and reviewing the literature. I'll let you know if I have any questions.

Glad to be of help!

[d-montgomery]

I ran the RegTests/EB-C12 with pelec.do_mol = 0 and pelec.plm_iorder = 2 as you suggested and can confirm the artificial gradient steepening. Here are the results prior to this PR:

and after this PR:

Thanks @magomedovs for catching this! I still need to verify the test failures but should be able to get this merged quickly.

[d-montgomery]

These EB test failures are expected as they all have do_mol = 0 with the default plm_iorder = 4. Whenever a cell is close enough to the EB surface where the 5-point stencil needed for 4th-order isn't fully available (but the 3-point stencil is), plm_iorder drops down to the 2nd order slope regardless of what plm_iorder was set to. Therefore, the diffs in these tests is a result of the fixed 2nd order slope.