Look for missing harmonic means in fluxes

[inducer]

When considering diffusive fluxes with coefficients "on the time derivative side" that can be discontinous, we should be using harmonic means. (As @majosm is heading towards with fluid/solid heat conduction). Look through our interior and boundary numerical fluxes where that's missing.

• For continuous coefficients, there will not be a difference.
• For "slightly" discontinuous coefficients, the difference might not be noticeable.
• For significantly discontinuous coefficients (which, given that we're doing DG, can always occur), this can be a pretty strong instability.

cc @MTCam @anderson2981 @lukeolson

[majosm]

A derivation of the harmonic mean for FV is here (search for "harmonic"): https://www.cae.tntech.edu/~shan/numerical%20heat%20transfer%20note(ME4730_5730)/1d-condution(2013).pdf

A quick attempt at justifying it for DG is here: https://notes.tiker.net/_jxEbQ5SSHSy24IKyQXjkQ