Adjoint phase adjustments

[smartalecH]

Closes #1507 and #1471.

• Updated the adjoint tests to include both single frequency and multi-frequency objective functions
• Fixed some DFT fields bugs within the multi-frequency case
• Rather than use the forward simulation's time profile, a Gaussian centered at the freq of interest is used. Interestingly, this seemed to fix the dependence of the forward simulation's pulse width (the test, for example, uses a bandwidth of 0.23*fcen).

Some caveats:

• We still need the adj_src_phase scale factor. This factor only accounts for the center frequency of the simulation. Is it possible that an additional phase factor is being added to the source by specifying the frequency parameter? (this parameter is used to determine at what frequency the mode solver should calculate the modes -- perhaps there's an additional phase term being added here..).
• The DFT field gradients aren't as accurate as they should be. Since we are planning to overhaul them in the same fashion as the Near2Far transform, I didn't spend much time debugging.
• The Eigenmode sources still only calculate the eigenmode at the center frequency, which introduces some error in the gradient. For the most part, it's a small order error.