Provide path to bindings for plugin

- Revert setHidden in menuitem_linux_gtk3.go: widgetSetVisible takes a
  `hidden` flag (true → gtk_widget_hide, false → gtk_widget_show) — the
  same convention as the GTK4 (linux_cgo.go) and purego (linux_purego.go)
  implementations. The original `widgetSetVisible(l.native, hidden)` was
  correct; the negation introduced in the previous commit inverted
  behaviour. Bug-bash item #8 in #5465 misread the function name.
- Make NewCalloc return *Calloc to match the pointer-receiver String/Free
  so the API is consistently pointer-semantic and chained calls like
  `NewCalloc().String(...)` continue to compile. Existing call sites
  (`c := NewCalloc(); defer c.Free()`) are unchanged.
- Update UNRELEASED_CHANGELOG.md accordingly: drop the (incorrect)
  setHidden line, note the NewCalloc return-type change.

Refs #5466 review (Copilot).

- Revert setHidden in menuitem_linux_gtk3.go: widgetSetVisible takes a
  `hidden` flag (true → gtk_widget_hide, false → gtk_widget_show) — the
  same convention as the GTK4 (linux_cgo.go) and purego (linux_purego.go)
  implementations. The original `widgetSetVisible(l.native, hidden)` was
  correct; the negation introduced in the previous commit inverted
  behaviour. Bug-bash item #8 in #5465 misread the function name.
- Make NewCalloc return *Calloc to match the pointer-receiver String/Free
  so the API is consistently pointer-semantic and chained calls like
  `NewCalloc().String(...)` continue to compile. Existing call sites
  (`c := NewCalloc(); defer c.Free()`) are unchanged.
- Update UNRELEASED_CHANGELOG.md accordingly: drop the (incorrect)
  setHidden line, note the NewCalloc return-type change.

Refs #5466 review (Copilot).

)

* fix(v3/linux/gtk3): resolve pre-existing bugs in legacy gtk3 path

Cluster of pre-existing bugs in the legacy GTK3 implementation files
(now `*_gtk3.go` after #5463). Surfaced by CodeRabbit on #5463; the
file rename made them visible but the bugs themselves predate the
default-flip. All affect the `-tags gtk3` build path only.

application_linux_gtk3.go
- run(): drop the early `return nil` on the file-association branch
  so startup handlers, common event setup, theme/power monitoring,
  and the GTK main loop still run after `ApplicationOpenedWithFile`
  is emitted. Restructure the surrounding `else` so the "not a URL,
  potential file open?" debug log no longer fires after a successful
  file-association match.
- getTheme(): guard `body[2]` access with `len(body) >= 3` and use
  comma-ok type assertions for `dbus.Variant` and the inner
  `string`, returning `("", false)` on any mismatch instead of
  panicking.

linux_cgo_gtk3.go
- appName(): drop the `defer C.free` on the pointer returned by
  `g_get_application_name()`. That memory belongs to GLib internal
  storage; `C.GoString` already copies into Go memory.
- clipboardGet(): nil-check the `gchar*` returned by
  `gtk_clipboard_wait_for_text` and `g_free` it after copying into
  a Go string. Previously leaked on every call.
- Calloc.String / Calloc.Free: change to pointer receivers so the
  `append(c.pool, ...)` actually grows the original pool. With value
  receivers the pool stayed empty and Free was a no-op, leaking
  every `c.String(...)` allocation.
- zoomIn / zoomOut: drop the negative `-1.10` multiplier (which
  `setZoom` then clamped to 1.0) and use the reciprocal of
  `zoomInFactor` for zoomOut. Promoted the factor to a package-level
  const and removed the stale "FIXME: assumed to be incorrect"
  comment.
- execJS: reuse the file-level `emptyWorldName` C string instead of
  allocating a fresh `C.CString("")` per call that was never freed.

menuitem_linux_gtk3.go
- setHidden: pass `!hidden` to `widgetSetVisible` (previously
  inverted; hiding a menu item showed it and vice versa).
- setAccelerator: drop the development `fmt.Println("setAccelerator", ...)`.

The identical `Calloc` value-receiver bug also exists verbatim in
`linux_cgo.go` (the now-default GTK4 path); intentionally left for a
follow-up to keep this PR scoped to the gtk3 tracking issue.

Refs #5465

* fix(v3/linux/gtk3): address Copilot review on #5466

- Revert setHidden in menuitem_linux_gtk3.go: widgetSetVisible takes a
  `hidden` flag (true → gtk_widget_hide, false → gtk_widget_show) — the
  same convention as the GTK4 (linux_cgo.go) and purego (linux_purego.go)
  implementations. The original `widgetSetVisible(l.native, hidden)` was
  correct; the negation introduced in the previous commit inverted
  behaviour. Bug-bash item #8 in #5465 misread the function name.
- Make NewCalloc return *Calloc to match the pointer-receiver String/Free
  so the API is consistently pointer-semantic and chained calls like
  `NewCalloc().String(...)` continue to compile. Existing call sites
  (`c := NewCalloc(); defer c.Free()`) are unchanged.
- Update UNRELEASED_CHANGELOG.md accordingly: drop the (incorrect)
  setHidden line, note the NewCalloc return-type change.

Refs #5466 review (Copilot).

* fix(v3/linux): apply Calloc value-receiver fix to the default GTK4 path

Mirror the same Calloc fix from linux_cgo_gtk3.go in linux_cgo.go (the
now-default GTK4 build). Without pointer receivers, `append(c.pool, ...)`
in `String` mutated a copy, so the pool stayed empty and `Free()` was a
no-op — every per-window `c.String(...)` allocation in
`setupSignalHandlers` and `windowNewWebview` leaked. With value
receivers gone, allocations are tracked and released on Free.

`NewCalloc` now returns `*Calloc` so the API is consistently
pointer-semantic across both build tags.

Refs #5466 review (Copilot).