fix(mpc): return error instead of panic when publicKey is nil in BroadcastPublicKey

registerReadyPairs spawns a goroutine that calls triggerECDHExchange ->
BroadcastPublicKey. If the peer restarts between scheduling and the
goroutine reading the local publicKey, e.publicKey is nil and
crypto/ecdh.(*PublicKey).Bytes panics with 'invalid memory address or
nil pointer dereference', killing the goroutine and forcing a
container restart+rejoin cycle of ~10-15s per occurrence.

Nil-check the receiver and e.publicKey at the top of BroadcastPublicKey
and return a descriptive error so the caller's ECDH retrigger logic
takes over instead of a panic propagating to stderr.

Refs fystack/mpcium issue 158.

Signed-off-by: SAY-5 <say.apm35@gmail.com>

Author: SAY-5

pkg/mpc/key_exchange_session.go

@@ -152,6 +152,13 @@ func (s *ecdhSession) Close() error {
 }
 
 func (e *ecdhSession) BroadcastPublicKey() error {
+	// Peer restart mid-handshake can schedule this broadcast goroutine
+	// before the local publicKey is populated; crypto/ecdh.(*PublicKey).Bytes
+	// then panics with nil pointer dereference. Return an error so the
+	// caller's retrigger logic takes over instead of the goroutine dying.
+	if e == nil || e.publicKey == nil {
+		return fmt.Errorf("ecdh session %s has no public key to broadcast yet", e.nodeID)
+	}
 	publicKeyBytes := e.publicKey.Bytes()
 	msg := types.ECDHMessage{
 		From:      e.nodeID,