Implement mortar correction procedure from Chan et al.

The projection mortar operator applied to the DG approximation with LGL nodes is stable for upwind surface fluxes but is unstable for central surface fluxes. This is due to the presence of positive eigenvalues in the operator. The recent mortar correction described by Jesse ([https://arxiv.org/abs/2005.03237](https://arxiv.org/abs/2005.03237)) fixes this issue, allowing one to use central fluxes at the mortars with LGL nodes.

No source file, test, or entry point is named. Start by reading the linked Chan et al. paper and locating the projection mortar operator and LGL-node handling; done means the mortar correction is implemented so central surface fluxes with LGL nodes no longer have the reported instability.

Written by the indexing model from the issue text.