[rogeriojorge]

Stacked on #66.

Adds magnetic=False and u_s=None to field_scaled_closure_rate. With magnetic=True the per-axis field scale becomes |E_i - <E_i>| + sum_{j != i} V_j |B_l - <B_l>| with V_j = |u_j| + a_j sqrt(2 N_j), the largest velocity the kept ladder on axis j represents. The default is the electric-only #66 rate, unchanged bit for bit.

Why: in Dark-SPECTRAX B09 (Weibel, 1D3V, k0 = 1.2), the adaptive step collapse at t ~ 110 is the AW truncation instability driven by v x B. With the fields frozen at the c = 0 state, the kinetic operator has a growing mode with Re lambda = 1.31 at t = 100 and 2.65 at t = 108, against the Weibel rate of 0.108. All of that mode's energy sits in the top third of the Hermite ladder. The electric-only rate leaves it almost unchanged (1.23 and 2.39). The magnetic scale brings it down to 0.0022 and 0.25. That moves the base-level collapse from t = 110 to 146, past the first magnetic-energy maximum (records and declared criteria: Dark-SPECTRAX studies/b09_closure.py, studies/b09/).

Test: analytic value with a non-uniform B_z and a drift, and an exact zero at c = 0.

[rogeriojorge]

Merged into integration/dark-baseline at dd80b0f (not into main).