Combo widget filter (control_after_generate) does not filter values on randomize, neither as plain substring nor as /regex/

### Custom Node Testing

- [x] I have tried disabling custom nodes and the issue persists (see [how to disable custom nodes](https://docs.comfy.org/troubleshooting/custom-node-issues#step-1%3A-test-with-all-custom-nodes-disabled) if you need help)

### Expected Behavior

When a filter value is set on a Primitive node's combo widget (e.g. `ckpt_name`), the "randomize" control_after_generate mode should only select values from the combo list that match the filter, whether typed as a plain substring (`ghost`) or as a regex pattern (`/ghost/`)

### Actual Behavior

The filter is ignored. After queuing the prompt / triggering randomize, the Primitive node selects a value from the full unfiltered combo list, including files that do not contain "ghost" in their name. Tested with both `ghost` and `/ghost/` as the filter value, same result in both cases.

### Steps to Reproduce

1. Add a CheckpointLoaderSimple node
2. Right-click the ckpt_name widget → Convert to Input
3. Add a Primitive node and connect it to the ckpt_name input
4. Set the Primitive's control_after_generate to "randomize"
5. Set the filter
6. Queue the prompt / trigger randomize multiple times
7. Observe the selected checkpoint does not respect the filter
8. Tested with --disable-all-custom-nodes to rule out custom node interference — issue persists

### Debug Logs

```powershell
[INFO] Adding extra search path checkpoints /home/rafaelfg/Documents/stable/ComfyUI/models/checkpoints
[INFO] Adding extra search path clip /home/rafaelfg/Documents/stable/ComfyUI/models/clip
[INFO] Adding extra search path clip_vision /home/rafaelfg/Documents/stable/ComfyUI/models/clip_vision
[INFO] Adding extra search path configs /home/rafaelfg/Documents/stable/ComfyUI/models/configs
[INFO] Adding extra search path controlnet /home/rafaelfg/Documents/stable/ComfyUI/models/controlnet
[INFO] Adding extra search path diffusion_models /home/rafaelfg/Documents/stable/ComfyUI/models/diffusion_models
[INFO] Adding extra search path diffusion_models /home/rafaelfg/Documents/stable/ComfyUI/models/unet
[INFO] Adding extra search path embeddings /home/rafaelfg/Documents/stable/ComfyUI/models/embeddings
[INFO] Adding extra search path loras /home/rafaelfg/Documents/stable/ComfyUI/models/loras
[INFO] Adding extra search path upscale_models /home/rafaelfg/Documents/stable/ComfyUI/models/upscale_models
[INFO] Adding extra search path vae /home/rafaelfg/Documents/stable/ComfyUI/models/vae
[WARNING] WARNING: You need pytorch with cu130 or higher to use optimized CUDA operations.
[INFO] Found comfy_kitchen backend cuda: {'available': True, 'disabled': True, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope_split_half', 'apply_rope_split_half1', 'convrot_w4a4_linear', 'dequantize_convrot_w4a4_weight', 'dequantize_int8_convrot_weight', 'dequantize_int8_convrot_weight_dtype', 'dequantize_int8_simple', 'dequantize_int8_simple_dtype', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'gemv_awq_w4a16', 'int8_linear', 'prepare_int4_weight_for_int8_linear', 'quantize_and_rotate_rowwise', 'quantize_convrot_w4a4_weight', 'quantize_int8_convrot_weight', 'quantize_int8_rowwise', 'quantize_int8_tensorwise', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'quantize_svdquant_w4a4', 'rms_rope', 'rms_rope1', 'rms_rope_split_half', 'rms_rope_split_half1', 'scaled_mm_nvfp4', 'scaled_mm_svdquant_w4a4', 'stochastic_rounding_fp8']}
[INFO] Found comfy_kitchen backend eager: {'available': True, 'disabled': False, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope_split_half', 'apply_rope_split_half1', 'convrot_w4a4_linear', 'dequantize_convrot_w4a4_weight', 'dequantize_int8_convrot_weight', 'dequantize_int8_convrot_weight_dtype', 'dequantize_int8_simple', 'dequantize_int8_simple_dtype', 'dequantize_mxfp8', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'gemv_awq_w4a16', 'int8_linear', 'prepare_int4_weight_for_int8_linear', 'quantize_and_rotate_rowwise', 'quantize_convrot_w4a4_weight', 'quantize_int8_convrot_weight', 'quantize_int8_rowwise', 'quantize_int8_tensorwise', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8', 'quantize_svdquant_w4a4', 'rms_rope', 'rms_rope1', 'rms_rope_split_half', 'rms_rope_split_half1', 'scaled_mm_mxfp8', 'scaled_mm_nvfp4', 'scaled_mm_svdquant_w4a4', 'stochastic_rounding_fp8']}
[INFO] Found comfy_kitchen backend triton: {'available': True, 'disabled': True, 'unavailable_reason': None, 'capabilities': ['adaln', 'apply_rope', 'apply_rope1', 'apply_rope_split_half', 'apply_rope_split_half1', 'dequantize_nvfp4', 'dequantize_per_tensor_fp8', 'int8_linear', 'quantize_and_rotate_rowwise', 'quantize_int8_rowwise', 'quantize_mxfp8', 'quantize_nvfp4', 'quantize_per_tensor_fp8']}
[INFO] Checkpoint files will always be loaded safely.
[INFO] Total VRAM 24107 MB, total RAM 64205 MB
[INFO] pytorch version: 2.9.1+cu128
[INFO] xformers version: 0.0.33.post2
[INFO] Set vram state to: NORMAL_VRAM
[INFO] Device: cuda:0 NVIDIA GeForce RTX 4090 : cudaMallocAsync
[INFO] Using async weight offloading with 2 streams
[INFO] Enabled pinned memory 57784.0
[INFO] working around nvidia conv3d memory bug.
[INFO] Using xformers attention
aimdo: /project/src-posix/cuda-funchooks.c:52:DEBUG:aimdo_setup_hooks: hooks successfully installed
aimdo: /project/src/control.c:247:INFO:comfy-aimdo inited for GPU: NVIDIA GeForce RTX 4090 (VRAM: 24106 MB)
[INFO] DynamicVRAM support detected and enabled
[INFO] Python version: 3.12.13 (main, Mar 6 2026, 11:39:28) [GCC 15.2.1 20260209]
[INFO] ComfyUI version: 0.29.2
[INFO] comfy-aimdo version: 0.4.10
[INFO] comfy-kitchen version: 0.2.22
[INFO] Initializing frontend: Comfy-Org/ComfyUI_frontend@1.49.3, requesting version details from GitHub...
[INFO] [Prompt Server] web root: /home/rafaelfg/Documents/stable/ComfyUI/web_custom_versions/Comfy-Org_ComfyUI_frontend/1.49.3
[INFO] Asset seeder disabled
[INFO] No OpenGL_accelerate module loaded: Acceleration disabled
[INFO] Skipping loading of custom nodes
[INFO] Context impl SQLiteImpl.
[INFO] Will assume non-transactional DDL.
[INFO] Using RAM pressure cache.
[INFO] Starting server

[INFO] To see the GUI go to: http://127.0.0.1:8188
```

### Other

_No response_

No source file or test is named. Start at the Primitive node's combo widget and its control_after_generate randomize path, reproduce with a filtered ckpt_name list using both `ghost` and `/ghost/`, and trace where the filter is applied. Done means repeated randomization selects only values matching the filter, with custom nodes disabled.

Written by the indexing model from the issue text.