solve_bicgstab: consistent RT casting

[eebasso]

Summary

This PR makes sure that every instance in solve_bicgstab of setting an RT variable to 0 or checking if it is equal to 0 uses the explicit RT(0.0) cast.

[WeiqunZhang]

Why do we need to change from

if (rhTv != RT(0.0))
{
    alpha = rho/rhTv;
}
else
{
    ret = 2; break;
}

to

if (rhTv == RT(0.0))
{
   ret = 2; break;
}
alpha = rho/rhTv;

I did not write the original code. Either way is fine with me. But why does it matter?

[eebasso]

This change takes up less lines of code, and makes this zero check consistent in structure with the rho == RT(0.0) and omega == RT(0.0) checks. In my humble opinion, it looks more elegant this way.

[asalmgren]

This is a small change for the sake of change; in the best interest of the developers' time we would like contributors to focus their efforts on fixing erroneous code (aka bugs) and contributing new functionality.