Old logic: check if instantaneous PV surplus > battery deficit (100% target).
Problem: moment of cloud = wrong decision; always aimed for 100% SOC.

New logic:
- Target SOC at 15:00 = what's needed to cover 15:00-22:00 consumption
  (expensive tariff) minus expected PV after 15:00, + 15% buffer
- Uses Solcast prognoza_na_nastepna_godzine * 1.8 to estimate PV in 13-15h window
- Uses Solcast pozostala_prognoza_na_dzis to estimate PV after 15:00
- Uses current load_power as average consumption proxy for 15-22h (7h)
- Grid charging current = exactly what's missing (not always 100A)
- program_4_soc = calculated target SOC (not always 100%)
- Co-15min optimization uses prognoza_na_biezaca_godzine + remaining window time

Result: charges battery to precisely what's needed for the day, not always 100%.
On sunny days: Battery First (no grid import). On cloudy days: minimal grid import.

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>