fix(sensor): 🐛 correct double /10 scaling on energy input sensors

[rhammen]

🐞 Problem

HEAT_ENERGY_INPUT, DHW_ENERGY_INPUT and COOLING_ENERGY_INPUT started reporting values 10x too low (e.g. 2699.38 kWh → 269.94 kWh), triggering "new cycle" resets in the recorder for total_increasing energy sensors.

🔍 Root cause

461875c registered parameters 1136/1137/1139 with the luxtronik library's Energy datatype, whose from_heatpump() already divides the raw register value by 10. The matching sensor descriptions still carried factor=0.01 — originally calibrated for the old Unknown (raw passthrough) datatype these parameters used before that commit. Combining Energy's /10 with the sensor's factor=0.01 (/100) gave a net /1000 instead of the intended /100, an extra 10x division.

🔧 Fix

• Change factor from 0.01 to 0.1 on the three affected sensor descriptions in sensor_entities_predefined.py, restoring the previously-correct scaling while keeping the more semantically correct Energy datatype.
• Correct the stale comment in lux_overrides.py that stated the wrong factor.
• Add TestEnergyInputScaling, which runs a raw register value through the real Energy datatype and the real production sensor description end-to-end, so a regression in either the datatype or the description factor is caught. Verified these tests fail (reproducing the reported 10x-low values) against the previous factor=0.01.

🧪 Testing

• pytest -q — 744 passed
• ruff check / ruff format --check — clean
• basedpyright — 0 errors

✅ Test plan

• Confirm sensor.luxtronik2_heat_energy_input, ..._dhw_energy_input, ..._cooling_energy_input report values consistent with pre-461875c readings after upgrading
• Confirm no further "new cycle" resets appear in the recorder for these sensors

🤖 Generated with Claude Code

[github-actions]

Package	Line Rate	Health
custom_components.luxtronik2	100%	✔
Summary	100% (2880 / 2880)	✔