Update boundary conditions before pressure solve

[glwagner]

No description provided.

[tomchor]

I don't see any difference in my simulations after this change. As an example, here's a headland simulation (similar to the paper one but with the original concave bottom) without this change, using an fft solver:

channel_16_fft_15.mp4

and the same simulation with this change:

channel_16.mp4

(Note that both simulations above also use the changes done in #4539)

[glwagner]

is there any change in the predictor velocity?

[glwagner]

The difference could be small, so I would look at the pointwise difference betwween the solutions. note that because we are talking about details associated with time-dependence, the difference in results should depend on the time-step, too.