Upgrade error estimation for time-dependent problems

[sebastiangrimberg]

Consider both electric flux density and magnetic field recovery error for transient problems. Note: The improved recovery is incorrect for "infinitely thin" interior boundaries applying a PEC or impedance BC. In this case these BCs break the continuity of the recovered H and D fields. This will be resolved in a future PR and is a separate improvement to the changes here.

TODO

• [x] Changelog
• [x] Rebaseline error indicators in (time dependent) examples

Note: Based on https://github.com/awslabs/palace/pull/237

[sebastiangrimberg]

Visualizations of the error indicator for the CPW example with lumped ports, at 60 GHz (to show the effect of multiple wavelengths):

• E-field magnitude at time = 0

• main:

• This PR:

We see that the correction for both energy terms (kinetic + potential, or magnetic + electric fields) makes the indicator more uniform over the length of the transmission line as we might expect, instead of preferentially indicating high errors at nodes of the electric field.