Consistent treatment of (effectively) shear-less case

[Ryusuke Numata]

Needs special treatment in several places if shat=0 or effectively shear-less (abs(shat)<1.e-5). In this issue, I’d like to make things correct for shear-less runs.

shat affects the wave numbers (kx, theta0, kperp2) in kt_grids, drift frequencies (wdrift, wcoriolis) in dist_fn, ExB shear and related diagnostics in dist_fn, and boundary condition (abs(shat) < 1.e-5 triggers periodic boundary), and maybe some more.

In most places abs(shat) < 1.e-5 means zero shear. Is this always true? Does this work fine?

[Ryusuke Numata]

Related PR #772, PR #781

[Ryusuke Numata]

Basically, the code relies on finite shat. Finite shat should always work fine. In other words, shat=0 is not basically acceptable.

For shat=0 case, we can define x and kx independent of ky and q and shat. It is natural to define x/L = (dpsi_N/drho) (rho-rho0). Then kperp2 = ky^2 |grad alpha|^2 + kx^2 |grad rho|^2. To use the existing code, we may define gds22 = (shat*|grad rho|)^2, which obviously works only for shat /= 0. (gds21 may be shat*grad alpha.grad rho.)

But, using shat parameter for shear-less case does not make sense to me.