fix: fix tailnet remoteCoordination to wait for server (#14666)

Fixes #12560

When gracefully disconnecting from the coordinator, we would send the Disconnect message and then close the dRPC stream.  However, closing the dRPC stream can cause the server not to process the Disconnect message, since we use the stream context in a `select` while sending it to the coordinator.

This is a product bug uncovered by the flake, and probably results in us failing graceful disconnect some minority of the time.

Instead, the `remoteCoordination` (and `inMemoryCoordination` for consistency) should send the Disconnect message and then wait for the coordinator to hang up (on some graceful disconnect timer, in the form of a context).

agent/agent.go

@@ -1357,7 +1357,7 @@ func (a *agent) runCoordinator(ctx context.Context, conn drpc.Conn, network *tai
 		defer close(errCh)
 		select {
 		case <-ctx.Done():
-			err := coordination.Close()
+			err := coordination.Close(a.hardCtx)
 			if err != nil {
 				a.logger.Warn(ctx, "failed to close remote coordination", slog.Error(err))
 			}

agent/agent_test.go

@@ -1896,7 +1896,9 @@ func TestAgent_UpdatedDERP(t *testing.T) {
 			coordinator, conn)
 		t.Cleanup(func() {
 			t.Logf("closing coordination %s", name)
-			err := coordination.Close()
+			cctx, ccancel := context.WithTimeout(testCtx, testutil.WaitShort)
+			defer ccancel()
+			err := coordination.Close(cctx)
 			if err != nil {
 				t.Logf("error closing in-memory coordination: %s", err.Error())
 			}
@@ -2384,7 +2386,9 @@ func setupAgent(t *testing.T, metadata agentsdk.Manifest, ptyTimeout time.Durati
 		clientID, metadata.AgentID,
 		coordinator, conn)
 	t.Cleanup(func() {
-		err := coordination.Close()
+		cctx, ccancel := context.WithTimeout(testCtx, testutil.WaitShort)
+		defer ccancel()
+		err := coordination.Close(cctx)
 		if err != nil {
 			t.Logf("error closing in-mem coordination: %s", err.Error())
 		}