fix(control): use the same battery power limits in planning and dispatch

[frahlg]

Problem and result

A 20 kWh battery with no configured power limits received a 10 kW planner limit (0.5C) while dispatch silently capped commands at 5 kW. The planner could schedule energy that Core would never command, even with the external ramp disabled.

Planning and dispatch now share one resolver. Missing limits use the existing 5 kW dispatch default; energy capacity does not establish a power rating. Driver limits and battery overrides retain their precedence. One-sided explicit zero still blocks that direction, and both-zero overrides retain the existing invalid-config fallback policy. Boot and config reload use the same resolver.

Scope and safety

• Removes duplicated limit resolution; no change to hardware commands, meter freshness, fuse guards, ramp settings or optimizer code.
• This does not raise the normal unconfigured dispatch limit to 10 kW or establish any device's actual rating. A both-zero invalid override now also uses the 5 kW fallback for missing ratings instead of deriving power from capacity.
• Checked open PRs and reviews. feat(modbus): share one TCP session per device and proxy it #1003 changes Modbus startup wiring and docs: keep one agent guide and correct the retired FTWDB note #1364 changes a retired flag's help text in main.go; these changes are confined to the two battery-limit builders, with no shared hunks. No existing PR branch was changed.
• Driver recovery behavior is in the separate companion PR fix(pixii): retain calibration fault while charge status is unknown device-drivers#147; neither fix claims to establish the cause of a particular device's low output.

Verification

• Regression tests parse battery configuration and check both planner limits and actual dispatch in both directions, with slew disabled. Cases cover missing limits, one known driver limit, battery overrides, one-sided zero and invalid both-zero overrides.
• The new regression fails in seven cases against an isolated copy of v0.138.0-beta.1 and passes with this fix. Existing configured-limit regressions also pass.
• make verify passes: Go suites, migration/container/release checks, native solver checks, vet and build.
• No UI change or hardware test. No release or deployment.

Checklist

• The change follows VISION.md and one selected scope.
• Checked overlapping PRs and documented the separate hunks above.
• Relevant checks cover changed behavior and failure paths.
• No UI changed.
• A Changeset is included.
• Every commit has a DCO sign-off.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-28T18:26:17.180842Z	fd534e9	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.