bridge: report what the chip's IrqMask kept out

[NomDeTom]

This moves vendor/virtual-sx1262 from 2861950 to f2604d1, the merge of MeshBench/virtual-sx1262#2, which is where VirtualSX1262::irqSuppressed() arrives. f2604d1 has the same tree as 8fb9623, the head of that PR, so the builds under Verification below cover this pin.

What this changes

The radio stats record (kRadioStats) grows from 37 bytes to 47:

bytes	field
0-36	unchanged
37-38	DIO1 mask (dio1Mask())
39-42	console baud rate: 0, since a native node's console is a socket
43-46	irqSuppressed(): IRQ events the firmware's IrqMask kept out of the status register

Bytes 37-42 follow the layout MeshBench already reads from emulated nodes, so the two writers send one shape. MeshBench reads the record by length; a host that predates bytes 37-46 ignores them. MeshBench/meshbench#745 reads bytes 43-46.

Verification

• With vendor/virtual-sx1262 at 2861950: build.sh simple_repeater fails with 'class VirtualSX1262' has no member named 'irqSuppressed'.
• With vendor/virtual-sx1262 at 8fb9623 (the head of Enforce the IRQ mask, time CAD scans, and decode the rest of the command set virtual-sx1262#2): build.sh simple_repeater builds against MeshCore companion-v1.17.0. The same bridge/main.cpp change, on a local branch pinned at 8fb9623, built simple_repeater and companion_radio against both companion-v1.16.0 and companion-v1.17.0.
• Those builds, run under MeshBench with Read the chip's IrqMask-suppressed count from native radio stats meshbench#745 and Record each node's radio counters in every experiment cell meshbench#746 applied (headless, fife-strict trimmed to three nodes, and scotland-strict), sent 47-byte records. MeshBench read the count: 1.16 nodes (IrqMask 0x0272) reported 24 and 1.17 nodes (0x0276) 12 on the three-node link.