Settle where an undiscoverable peripheral gets configured, and when its applet button is visible

Follow-up from the review of #4905 (review, item 2). This is a maintainer preference call, which is why it is an issue rather than a change requested on either PR.

The divergence

Two peripherals landed in the same week with opposite answers to the same question:

Why neither is wrong on its own

The GHE justification checks out. AG/SS visibility really is driven by live discovery (AntennaGeniusModel::deviceDiscovered → setAgVisible/setShackSwitchVisible, MainWindow.cpp:2066-2070), so with no Everyware discovery path there is genuinely nothing to condition on — and an operator cannot type an address into a tile they cannot open. KSDR is the existing precedent for an always-visible, undiscoverable accessory, and it is a fair one.

The cost is that every AetherSDR user gets a "GHE" tile in the applet bar for an accessory almost none of them own.

A middle path neither PR takes

Condition the button on a stored host rather than on discovery — hidden until configured, shown thereafter — with first-time configuration living wherever the project decides accessory setup belongs (the RadioSetupDialog Peripherals tab being the obvious candidate, per #4919).

The ask

Not a change to either PR. A decision on the convention, so that peripheral number five does not have to guess. Divergence is cheap to settle at two and expensive at five.

cc @ten9876

Start by reading the review of #4905 and compare the existing GHE and VK3AMP arrangements, including MainWindow.cpp:2066-2070 and the RadioSetupDialog Peripherals tab. Decide and document the project convention for undiscoverable peripherals, including whether applet buttons depend on a stored host or remain always visible and where first-time configuration belongs. Done means the convention is explicit for future peripheral work.