[rogeriojorge]

Adds remap_trigger, which decides a remap per species from the measured moments.

When it fires:

• |U - u|/a > 0.3;
• or |a'/a - 1| > 0.05 (a' capped);
• or, optionally, the tail fraction above 2N/3 exceeds tail_on, has doubled since the last event, and the move is non-trivial.

A remap resets the moment measures to round-off, so firing again needs the same growth (hysteresis). Species that do not fire keep their basis. For gradients, record the events and replay them as a frozen schedule with remap.

Tested SHA 045abc1. Command: python -m pytest -q -n 2 tests. Result: 179 passed.

New tests:

• Manufactured shifted and heated Maxwellians fire or not as expected, and never fire again after their own remap.
• A slow drift of 0.05 a per check fires every 6 or more checks, never on consecutive checks.
• The tail trigger needs both growth and a move.
• A live segmented plasma-oscillation run with at least 2 events:
  • its recorded schedule replays bitwise;
  • jax.grad through the replay matches central finite differences to 1e-6 relative.

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.