Exploit diagonalness of the mass matrix when applying tensor product mass inverse

[inducer]

When?

• If using (tensor product) Gauss-Legendre points,
• or if using Lobatto-Gauss-Legendre points with permission to lump (Interface for LGL mass lumping? #348).

This is related to #349, but for the inverse operator.

Sample reference: https://doi.org/10.1002/fld.3923

cc @MTCam @a-alveyblanc @lukeolson

[a-alveyblanc]

Shouldn't the mass/mass inverse already be diagonal if we are using a nodal basis?

[inducer]

Not for LGL points, see #343.

[a-alveyblanc]

I think you mean #344