Moving Hermite basis 2/5: exact moving-basis terms and the fused pump frame

[rogeriojorge]

Adds the exact rates for a moving and rescaled basis, and an oscillating-centre (pump) frame.

basis_rate_terms. Per axis, with C = n* g/(a_x a_y a_z):

-(u_dot/a) sqrt(2n) C_{n-1} - (a_dot/a)[(n+1) C_n + sqrt(n(n-1)) C_{n-2}]

• In g this is -(a_dot/a)[n g_n + ...]. The +1 comes from the 1/a normalization.
• Both terms only lower the index, so no closure is needed.

Pump frame.

• uniform_acceleration returns u_dot = (q/m)(E0 + u x B0) and the uniform field F0.
• Hermite_Fourier_system(..., F0=F0) uses E - E0 and u x (B - B0), so the shift and the uniform force cancel identically.
• simulation(frame="pump") moves each centre this way. Width changes go through the remap (PR 3/5), not a continuous a_dot.
• F0=None (the default) is bitwise the previous kernel.

Tested SHA 39987bf. Command: python -m pytest -q -n 2 tests. Result: 158 passed.

New tests:

• basis_rate_terms matches 4th-order finite differences of exact quadrature projections of a non-Gaussian 3V f onto a moving, rescaled basis (all three axes, unequal orders) to 1e-10 relative.
• The fused force equals the full force plus the shift term to 1e-12 (random 3-species state, uniform E and B).
• Uniform electron-ion plasma oscillation in the pump frame to t = 1000:
  • g stays delta_n0 for both species to 1e-13;
  • the centres follow the analytic two-fluid solution to 1e-8;
  • total energy is conserved to 3e-9 at tolerance 1e-12, with 3.6k steps. The same run in a fixed basis needs 22.5k steps and drifts 2e-11.

Stack: base adaptive/base = integration/dark-baseline 9d0982d (main ab87385 + #48, #46, #13, #12, #9, #55, #50, #44, #18, #51). This lands on main after those foundation PRs and the earlier PRs of this stack (#56 first). Local results only (macOS, Python 3.12, JAX 0.11.2, x64); GitHub CI on this branch has not been checked yet.