[Bug]: Fleet plan includes offers from backends without multi-node support with placement: cluster

[r4victor]

Steps to reproduce

1. Create a fleet configuration with placement: cluster.
2. Apply configuration.
3. Notice offers from clouds that do not support multi-node:

✗ dstack apply -f .dstack/confs/fleet.yaml            
 Project        r4victor-main                          
 User           victor                                 
 Configuration  .dstack/confs/fleet.yaml               
 Type           fleet                                  
 Fleet type     cloud                                  
 Nodes          2                                      
 Placement      cluster                                
 Resources      2..xCPU, 8GB.., 8xa100, 100GB.. (disk) 
 Spot policy    on-demand                              

 #  BACKEND  REGION            INSTANCE     RESOURCES        SPOT  PRICE      
 1  lambda   europe-central-1  gpu_8x_a100  124xCPU,         no    $10.32     
                                            1933GB, 8xA100                    
                                            (40GB),                           
                                            6598.7GB (disk)                   
 2  lambda   asia-northeast-1  gpu_8x_a100  124xCPU,         no    $10.32     
                                            1933GB, 8xA100                    
                                            (40GB),                           
                                            6598.7GB (disk)                   
 3  vastai   us-washington     17831846     96xCPU, 1814GB,  no    $13.8944   
                                            8xA100 (80GB),                    
                                            100.0GB (disk)                    
    ...                                                                       
 Shown 3 of 37 offers, $51.2073 max

Specifying 8xa100 helps getting backends without multi-node support on top.

Actual behaviour

This is caused by the multinode check that relies on fleet_model that does not exist when requesting the plan:

dstack/src/dstack/_internal/server/services/fleets.py

Line 288 in c4ddee1

multinode = fleet.spec.configuration.placement == InstanceGroupPlacement.CLUSTER

Note that the offers without multi-node are actually filtered out when provisioning the fleet, so it's only a fleet plan bug.

Expected behaviour

No response

dstack version

master

Server logs

Additional information

No response