Add hybrid resistivity current term

[clarkse]

This PR adds a current magnitude term in the hybrid resistivity to allow for scaling of the resistivity as a function of the magnitude of total J.

[roelof-groenewald]

Suggested change

	Real jx_val = Interp(Jx, Jx_stag, Ex_stag, coarsen, i, j, k, 0);
	Real jy_val = Interp(Jy, Jy_stag, Ex_stag, coarsen, i, j, k, 0);
	Real jz_val = Interp(Jz, Jz_stag, Ex_stag, coarsen, i, j, k, 0);
	Real jx_val = Interp(Jx, Jx_stag, Ez_stag, coarsen, i, j, k, 0);
	Real jy_val = Interp(Jy, Jy_stag, Ez_stag, coarsen, i, j, k, 0);
	Real jz_val = Interp(Jz, Jz_stag, Ez_stag, coarsen, i, j, k, 0);

[clarkse]

Thanks for catching the staggering bug.

[roelof-groenewald]

I think the preferred way to calculate it would be:

Suggested change

	jtot_val = sqrt(pow(jr_val, 2) + pow(jt_val, 2) + pow(jz_val,2));
	jtot_val = sqrt(jr_val*jr_val + jt_val*jt_val + jz_val*jz_val);

and same for the other directions and Cartesian case.

[clarkse]

I have dropped using the pow call. Just curious though, with compiler optimization wouldn't pow(,2) be inlined anyway?

[roelof-groenewald]

Yes, with new enough compilers it should be but the reasoning is that with a simple square, the x*x approach will never be less efficient than pow(x, 2) whereas the latter could be less efficient with sub-optimal compiler flags.

[roelof-groenewald]

So, actually with the change I don't think you need the static_cast<Real> anymore if you use std::sqrt, even in a lambda function now (#4490).

[roelof-groenewald]

Thanks for the check to only calculate j_tot if it will be used.

[roelof-groenewald]

Thanks! I'll be curious to hear how you are using the current dependent resistivity.