Fix add_tiny_price_slope for flat prices to reliably break ties in LP optimizer

[Copilot]

With constant energy prices, the tie-breaking slope added by add_tiny_price_slope was too small for the LP solver to distinguish early from late charging, causing charging to scatter across time slots instead of front-loading as expected.

Root cause

When price_spread == 0, the fallback penalty was a fixed absolute value max_penalty = d = 1e-4 (0.0001 EUR/MWh). At 100 EUR/MWh flat prices this creates a total slope of 0.0001 EUR/MWh across the scheduling window — within HiGHS's numerical tolerance, so the solver treated all time slots as equivalent.

Fix

Scale the fallback by the actual price level instead of a fixed constant:

# Before
else:
    max_penalty = d  # always 1e-4, too small for e.g. 100 EUR/MWh flat prices

# After
else:
    max_penalty = max(abs(prices[col_name].mean()), 1.0) * d  # 0.01 EUR/MWh at 100 EUR/MWh

This produces a 100× larger slope (0.01 EUR/MWh) for typical energy prices, reliably above LP solver tolerance.

Test assertions added

test_two_flexible_assets_with_commodity now verifies the expected early-charging behaviour and electricity costs:

• Battery (20 kW cap, SOC 20→80 kWh, η=0.95): draws 60/0.95 ≈ 63.16 kWh → 3 full slots at 20 kW + partial slot, cost ≈ 6.316 EUR
• Heat pump (10 kW cap, SOC 10→40 kWh, η=0.95): draws 30/0.95 ≈ 31.58 kWh → 3 full slots at 10 kW + partial slot, cost ≈ 3.158 EUR
• Both devices: zero power in slots 4–23

---

✨ Let Copilot coding agent set things up for you — coding agent works faster and does higher quality work when set up for your repo.