Loadpoint: split sense and control, parallel maintainance

[premultiply]

Main part of the loadpoint sense/control epic #30366.

• Fast, independent sensing — each loadpoint senses status + measurements on its own 2s loop, so a "vehicle plugged in" is detected within seconds regardless of how many loadpoints the site has (no longer N × interval).
• Event-driven control — control runs when something actually changes (fairly scheduled so no loadpoint starves), instead of one loadpoint per interval tick.
• Parallel actuation — loadpoints no longer wait for each other. A self-correcting in-flight reserve (site surplus + per-circuit headroom) hides power a loadpoint just claimed until the meters catch up, so parallel ramps can't double-allocate surplus or overshoot a circuit. The reserve is valid for one control interval and collapses the instant the meter reflects the new draw.

[andig]

each loadpoint senses status + measurements on its own ~2s loop

Do they run concurrently? We should create/check race builds if this surfaces hidden bugs.

And if we do this- how do we treat cloud apis? Meters/ chargers are not prepared for this request volume.

[premultiply]

each loadpoint senses status + measurements on its own ~2s loop

Do they run concurrently?

Yes.

how do we treat cloud apis? Meters/ chargers are not prepared for this request volume.

This is #30375