feat(panel): jump to top/bottom with cmd+up/down

[StuBehan]

Summary

Adds ⌘↑ / ⌘↓ to jump the current page to its first / last item, from any
scroll position — on every list page in the panel.

Changes

• One ⌘↑/↓ case in the existing cmdOnly block of panelHandlesKey,
  dispatching per mode through a new jumpToEdge(top:):
  • Events / Sessions / Tickets / Settings / Phrases move the selection to
    the first/last row — the viewport follows, since each page already scrolls
    to its selection.
  • Usage — the client list jumps to first/last client; the detail pane
    (no selection) scrolls the clip view to the top/bottom edge.
  • Modal / single-purpose screens (bootstrap, uninstall, update,
    post-update) return false, so the keystroke passes through untouched.
• selectFirst/selectLast helpers on EventStore, Phrases, and PanelNav
  (settings rows, outcomes index, usage client), plus a scrollDetailToEdge
  analog to the existing scrollDetailBy.
• Each page footer gains a Top/Bottom ⌘↑↓ hint.

Testing

• ./build.sh → Build complete, zero warnings (rebased onto 1.20.3).
• EventStoreTests and SettingsRowTests cover the first/last jump helpers
  (incl. empty-store); run in CI via swift test.
• Needs a manual eyeball (the panel can't be driven in CI): ⌘↑/↓ on each tab —
  lands on the first/last row, and the Usage detail pane scrolls to its edges.

Related issues

Companion to the Tickets-tab cursor-nav smoothing (#113). No tracked issue.

[hiskudin]

Reviewed the diff. Small, focused, tested — no changes needed from me.

• Dispatch table in jumpToEdge is the right shape: .sessions guards renamingPID == nil before intercepting (so ⌘↑/↓ passes through into the rename field where it's expected as text-nav), and modal/single-purpose screens fall through the default returning false.
• Usage split (usageDetailFocused → scrollDetailToEdge, otherwise client list first/last) mirrors the existing ↑↓ handling. scrollDetailToEdge clamps maxY = max(0, doc.frame.height - clip.bounds.height) correctly for the short-content case.
• Empty-store test on EventStore.selectFirst is a nice touch — asserts selectedID == nil rather than crashing.

Tiny style nit only (not worth a change): selectFirstUsageClient / selectLastUsageClient are semicolon-chained one-liners while selectFirstRow / selectLastRow a few functions later use guarded assignment across lines. Trivially inconsistent.

Ship it.