fix(driver): ferroamp aggregates battery across N ESOs

[erikarenhill]

Summary

• ferroamp.lua cached only the most-recently-published ESO in a single eso_data slot, so the battery power readout was the per-unit number instead of the cluster aggregate. On a 2×ESO site this halved bat_w; the controller's grid-chase loop fed on the wrong number and dispatch stuck ~190 W of import against a 0 W target.
• Replace the single slot with eso_data_by_id keyed on each ESO's id field. Sum DC power across all live ESOs (cell-monitor ubat × ibat), average voltage / SoC / DC-link, sum currents and Wh counters, surface per-ESO faults via worst-of relay/faultcode.
• Backward-compatible: messages without an id field land under a synthetic key, so single-ESO firmware and sims keep behaving identically.
• New eso_count diagnostic metric so operators can see how many ESOs the driver is summing.

How it was discovered

Live debugging session on a 2×ESO (21030026 + 23010216) Ferroamp install:

Source	Battery output
42 bat_w	153 W
Ferroamp ehub.pbat (aggregate DC)	306 W
Σ per-ESO ubat × ibat	305.8 W

After the fix the same site shows ratio 1.00 between 42 and ehub.pbat, and the grid chase converges normally.

Test plan

• go test ./internal/drivers/ -run Ferroamp -v — 8 existing tests + 2 new (TestFerroampDriverSumsBatteryPowerAcrossMultipleESOs, TestFerroampDriverSurfacesPerESOFaultsWorstOf)
• go test ./... — full suite green
• Deployed to a live 2×ESO site; bat_w matches ehub.pbat 1:1 over multiple snapshots; dispatch target tracks delivered discharge; grid converged inside grid_tolerance_w.

🤖 Generated with Claude Code

[erikarenhill]

@frahlg please have a test-drive with this driver in your system too and merge if its working. Tested on a ferroamp site with 2 ESO's which fixed lots of incorrect reportings