feat: add SOC kWh fallback for vehicles without realtimePower

[nanomad]

Summary

• For vehicles (e.g. MGS5) where RvsChargeStatus.realtimePower is always None or 0, DRIVETRAIN_SOC_KWH was silently never published
• Adds extract_soc_kwh(charge_status, soc) to the extractors module, following the same pattern as extract_soc / extract_electric_range
• Prefers the realtimePower-derived value when valid (> 0); falls back to soc% × real_total_battery_capacity otherwise
• Publishing moves from RvsChargeStatusPublisher to update_data_conflicting_in_vehicle_and_bms, where both the charge result and the resolved SoC% are already in scope

Design notes

• soc_kwh is now carried through RvsChargeStatusProcessingResult → ChrgMgmtDataRespProcessingResult as a computed field (no mutable state, no wiring through call sites)
• __validate_and_convert_soc_kwh enforces the > 0 rule for both the primary reading and real_total_battery_capacity, consistent with the existing validator pattern
• The fallback activates from the second poll onward on first startup (SoC% from extract_soc is set during update_data_conflicting_in_vehicle_and_bms, which runs after charge status is published)

Test plan

• tests/status_publisher/test_rvs_charge_status.py — unit tests for RvsChargeStatusPublisher: soc_kwh present when realtimePower is set, None when zero or missing
• tests/test_extractors.py — unit tests for extract_soc_kwh: primary path, fallback path, all None/zero edge cases
• tests/test_vehicle_handler.py — integration test for full polling cycle publishing DRIVETRAIN_SOC_KWH, and fallback test with realtimePower=None
• Full suite: poetry run pytest tests/ — 166 passed

🤖 Generated with Claude Code