[salmun-nister]

Fixes #16

Display-side fix for the static ALL-TIME row (option 2 in the issue): the daemon keeps allTime as completed days only, so the panel adds today's bytes before formatting.

Model.js - null-safe helper next to fmtGB:

// Lifetime usage for display. The daemon keeps `allTime` as completed days
// only (rolled at local midnight), so today's bytes are added here: otherwise
// ALL-TIME sits unchanged all day and can read below TODAY.
function lifetimeGB(allTimeGB, todayGB) {
  var a = num(allTimeGB)
  var t = num(todayGB)
  if (a === null && t === null) return null
  return (a === null ? 0 : a) + (t === null ? 0 : t)
}

Panel.qml - the usage row:

-                ? "ALL-TIME ↓ " + Model.fmtGB(s.usage.allTimeDownGB) + " ↑ " + Model.fmtGB(s.usage.allTimeUpGB)
+                ? "ALL-TIME ↓ " + Model.fmtGB(Model.lifetimeGB(s.usage.allTimeDownGB, s.usage.downGB))
+                  + " ↑ " + Model.fmtGB(Model.lifetimeGB(s.usage.allTimeUpGB, s.usage.upGB))

Testing

case	expected
both counters present	lifetime equals ledger + today (e.g. 10.8 + 3.2 -> 14.0 GB)
no ledger yet (fresh install)	lifetime equals today, not 0.0 GB
today missing	lifetime equals the ledger
both missing	— (null, not 0.0 GB)
after the midnight roll	ledger absorbs today and today resets to zero, so the displayed total is unchanged

Layout is unchanged - the string keeps the same shape and still formats through fmtGB.

Scope

Model.js and Panel.qml only. No Go changes, so backend/bin/starlinkd is untouched and the release gate's byte-identical check is unaffected. The rows this touches are separate from the ones in the power-block and 24 h-chart PRs, so it applies independently.

Known trade-off

status.json still reports the rolled figure (usage.allTime*), so the panel and the file disagree by one day's bytes. Moving the sum into the status builders is the follow-up described in the issue.