fix: scope EVC cumulative-energy big word order to GEN1 only

[Bl00d-B0b]

Fixes the word-order half of #2230; follow-up to the discussion on #2178.

The order32="big" override on charge_added_cum was hardware-verified on my X3-EVC-11kW Gen1 (serial C31103*, ARM v1.18, register 0x25 = 118, classified GEN1 by the plugin): at the same instant, the input copy 0x10-0x11 read [0, 3] (high-word-first, 0.3 kWh only when decoded big) while its holding twin 0x619-0x61A read [3, 0] and decoded to the correct 0.3 kWh little-endian. On that unit the documented order is demonstrably wrong for the input copy.

#2230 shows the opposite population: an X3-HAC 22kW Gen2 (ARM v2.03) follows the documented little order, so the global override inflated its lifetime total by exactly a factor of 65536 (322.6 kWh became 21,141,913.6 - the reporter's cross-check against the independent HTTP integration makes it airtight).

Fix: the definition is split per generation - GEN1 keeps order32="big" (hardware-verified), GEN2 returns to the documented default order. If more firmware populations surface (e.g. #2213, where neither order looks plausible), the same split can be refined per firmware version.

[TCWORLD]

I'd already opened a PR to fix this.

[wills106]

Does this need closing down then?

[Bl00d-B0b]

@wills106 yes - closing. #2234 implements the same generational split (and covers GEN3/GEN4 explicitly, which is the better version) and is already merged; I opened this one before spotting TCWORLD's. Verified #2234 against my Gen1 unit's case: GEN1 keeps the big-endian decode it needs, the HAC Gen2 reports in #2230 get the documented order. Nothing from this PR remains needed.