Removes manual grid manipulation in VerticallyStretchedRectilinearGrid constructor

[glwagner]

These lines should not be there and at the least lower the accuracy of calculations on stretched grids.

We need tests that the metrics and grid geometry are correct for stretched grids (eg all the cell spacings added up gives the total domain size; cell centers are located halfway between cell interfaces, etc).

Not sure why these lines are there but it may have been necessary due to a serious bug with the vertically stretched Poisson solver that was fixed in #1541. I'm not sure.

[francispoulin]

This all sounds good.

Would it also be worthwhile having a test for the finite difference operator to make sure that the accuracy is correct everywhere?

[glwagner]

The difference operator doesn't use the grid metrics --- but maybe you mean the derivative / gradient operator?

Such a test would catch this issue, it's true. We could use a simple test with linear gradients. There might already be such a test, but not for all grids... ?

I think the fact that the cell spacings don't add up to the grid size could be an even worse problem because it means that integrals / budgets are not correct. But I'm not sure since the spacings at cell centers might be more important (these lines only manipulate the spacing at cell interfaces).

[glwagner]

Tests pass so I guess we don't need these lines.

[navidcy]

please bump up minor version