…ethersdr#3383). Principle VIII. (aethersdr#3393)

## Summary

Fixes aethersdr#3383. Toggling DIV off now collapses the ESC panel space
immediately. The three sites that toggle `m_escPanel` visibility called
`resize(sizeHint())` right after `setVisible()`, but `setVisible()` only
posts a `LayoutRequest` to the audio tab's layout - it does not activate
it synchronously - so `sizeHint()` still returned the stale pre-hide
height and the panel kept a blank gap until something else (like
clicking the speaker sub-tab) forced a reflow. `adjustSize()` activates
the pending layout before computing the new size, which is the missing
step. All three diversity-path call sites in `src/gui/VfoWidget.cpp` are
swapped: the DIV button `toggled` handler, `setDiversityAllowed(false)`
(non-dual-SCU reconnect path), and the `SliceModel::diversityChanged`
model-sync handler. The DIV-on path benefits equally: the panel grows in
one step instead of relying on a previously-posted layout pass.

## Context

aethersdr#3383 reports the gap between the Med (AGC speed) row and the bottom of
the panel after DIV is toggled off, with the layout only correcting
after a tab switch. The @aethersdr-agent triage on the issue traced it
to exactly these three call sites, and this PR follows that triage. The
fourth `resize(sizeHint())` in the RADE info-row toggle is intentionally
left as-is to keep this scoped to the diversity paths.

## Constitution principle honored

Principle VIII - Evidence Over Assertion. Root cause and call sites come
from the @aethersdr-agent triage on aethersdr#3383 plus Qt's documented
`LayoutRequest`/`activate()` ordering (`QWidget::adjustSize()` activates
the layout before reading the size hint). The diff is confined to those
three sites.

## Test plan

- [x] Reproduction steps documented if user-reported bug (repro and
expected behavior from aethersdr#3383: DIV off collapses the AF/SQL/Med rows
immediately; DIV on restores the full ESC panel height; rapid toggling
settles correctly since every toggle now activates the layout before
sizing)
- [x] Existing tests pass (CI) - no test code touched; the CI build
matrix exercises compilation of the two-method swap

Not run against a real radio: no FlexLib or protocol surface is touched,
and this machine has no Qt6 toolchain for a local GUI build, so
compilation is delegated to CI.

## Checklist

- [x] Commits are signed (`docs/COMMIT-SIGNING.md`)
- [x] No new flat-key `AppSettings` calls - none added (Principle V)
- [x] All meter UI uses `MeterSmoother` - no meter changes (Principle
II)
- [x] Documentation updated if user-visible behavior changed - the
visible change is the bug fix itself; no docs reference the stale-gap
behavior
- [x] Security-sensitive changes reference a GHSA if applicable - not
security-sensitive

Co-authored-by: Matt Van Horn <455140+mvanhorn@users.noreply.github.com>