ModelPatcherDynamic: force load non leaf weights

[rattus128]

The current behaviour of the default ModelPatcher is to .to a model only if its fully loaded, which is how random non-leaf weights get loaded in non-LowVRAM conditions.

The however means they never get loaded in dynamic_vram. In the dynamic_vram case, force load them to the GPU.

Example test conditions:

RTX5090, Linux, ryzen 9600 underclocked to 2GHZ, --fast dynamic_vram, LTX2 800x600x121f

Before:

Requested to load LTXAV
Model LTXAV prepared for dynamic VRAM loading. 20542MB Staged. 0 patches attached.
100%|██████████| 20/20 [00:13<00:00,  1.50it/s]                                  
Model VideoVAE prepared for dynamic VRAM loading. 2331MB Staged. 0 patches attached.
lora key not loaded: text_embedding_projection.aggregate_embed.lora_A.weight
lora key not loaded: text_embedding_projection.aggregate_embed.lora_B.weight
Requested to load LTXAV
0 models unloaded.
Model LTXAV prepared for dynamic VRAM loading. 20542MB Staged. 1370 patches attached.
100%|██████████| 3/3 [00:03<00:00,  1.13s/it]                                   
Requested to load AudioVAE
loaded completely; 484.30 MB usable, 415.20 MB loaded, full load: True
0 models unloaded.
Model VideoVAE prepared for dynamic VRAM loading. 2331MB Staged. 0 patches attached.
Prompt executed in 36.05 seconds

After:

Requested to load LTXAV
Model LTXAV prepared for dynamic VRAM loading. 20542MB Staged. 0 patches attached.
100%|██████████| 20/20 [00:09<00:00,  2.17it/s]                                  
Model VideoVAE prepared for dynamic VRAM loading. 2331MB Staged. 0 patches attached.
lora key not loaded: text_embedding_projection.aggregate_embed.lora_A.weight
lora key not loaded: text_embedding_projection.aggregate_embed.lora_B.weight
Requested to load LTXAV
0 models unloaded.
Model LTXAV prepared for dynamic VRAM loading. 20542MB Staged. 1370 patches attached.
100%|██████████| 3/3 [00:02<00:00,  1.01it/s]                                   
Requested to load AudioVAE
loaded completely; 461.44 MB usable, 415.20 MB loaded, full load: True
0 models unloaded.
Model VideoVAE prepared for dynamic VRAM loading. 2331MB Staged. 0 patches attached.
Prompt executed in 31.76 seconds