fix(discovery): survive a peer disconnecting mid-session

[XxHugheadxX]

What changes

The scanner no longer dies when a beacon it discovered goes away.

DiscoveryDHT attached an error handler to the raw Hyperswarm connection but
not to the FramedStream wrapping it. When a peer disconnected, the stream
emitted an unhandled 'error' and took the whole process down:

Uncaught Error: connection reset by peer
    at WritableState.afterDestroy (node_modules/streamx/index.js:572:19)
    at FramedStream._destroy (node_modules/streamx/index.js:725:5)
  code: 'ECONNRESET'

One line: frames.on('error', () => {}), alongside the handler the connection
already had. A peer going away is normal in P2P, not exceptional.

Why

peer-lost already worked — the scanner removed the beacon and rendered
"No beacons nearby yet...". Then it crashed a moment later, so in practice a
traveler's session ended every time a shop closed its terminal.

Verified

Two instances with separate storage, full cycle:

• Scanner discovers the beacon — ID: 0550aef8749b...
• Beacon killed → scanner drops it, renders "No beacons nearby yet..."
• Scanner stays alive — no ECONNRESET, process still running
• Beacon restarted → scanner rediscovers it — ID: bbddb59bf01f...
• pnpm run lint — exit 0
• pnpm test — 4/4 tests, 5/5 asserts

Not addressed here

A restarted beacon comes back with a new id, because getStorageDir() in
backend/index.js suffixes the corestore path with process.pid. The same shop
reads as a different peer after every restart, and storage grows without bound.
Worth a separate change.