Fix stale charge-limit threshold in battery status panel

[jastincheis]

Problem

omarchy-battery-status preferred UPower's charge-*-threshold fields over the device's own sysfs value. The macsmc-battery driver never calls power_supply_changed() for a threshold-only write, so UPower's cached copy goes stale the moment battery-charge-limit changes it and never catches up.

Symptom: after raising the charge limit via the power panel's slider (e.g. 80% → 87%), the panel kept reporting/holding at the old limit (80%) — even though the real hardware threshold in sysfs was applied correctly and the battery would in fact charge past 80%.

Panel.qml already worked around this exact staleness for the slider's own display, by reading battery-charge-limit status (which reads sysfs) instead of UPower:

// Read straight from the CLI (which reads sysfs directly) rather than from
// UPower's cached charge-end-threshold: the macsmc-battery driver never
// calls power_supply_changed() for a threshold-only write, so UPower's
// copy goes stale the moment you change it and never catches up.

omarchy-battery-status never got the equivalent fix — it still checked UPower's value first and only fell back to sysfs when UPower returned nothing, which almost never happens (UPower always returns some — just possibly stale — value).

Fix

Swap the read order in omarchy-battery-status: read the device's own charge_control_{end,start}_threshold from sysfs first, falling back to UPower's cached fields only when sysfs isn't readable, matching the approach already used in Panel.qml.

Verification

• bash test/shell.d/battery-status-test.sh — passes
• ./test/shell — same 233 files, same 5 pre-existing failures as on quattro before this change (confirmed via git stash); none touch battery status
• Live check on an Apple Silicon machine with the limit set to 87% via the slider:
  • before: threshold 75-80% (stale UPower value)
  • after: threshold 82-87% (live sysfs value)

🤖 Generated with Claude Code

[malik-na]

Please add a focused regression proving sysfs wins when UPower returns a different, non-empty threshold. That stale-cache case is the bug, and this PR currently changes production code without changing its test.