Add circular tiling support (for making seamless textures)

[Phylliida]

Adds additional functions:

• ggml_conv_2d_circular
• ggml_conv_2d_dw_direct_circular
• ggml_conv_2d_direct_circular
• ggml_pad_circular

These tweak the inputs to the shaders (which are modified in cuda, cpp, and vulkan) to cause them to behave in a "circular" way (like on a torus, the boundaries wrap around).

This is useful for making seamless textures, for example here is a texture generated using this command in stable-diffusion-cpp:

./bin/sd  --diffusion-model /home/bepis/Documents/world/models/qwen-image-lighting-8steps-V1.0-Q4_K_S.gguf  --vae /home/bepis/Documents/world/models/qwen_image_vae.safetensors  --qwen2vl /home/bepis/Documents/world/models/Qwen2.5-VL-7B-Instruct-IQ4_XS.gguf --cfg-scale 2.5 --sampling-method euler -p "onion" --circular --seed 420

By default this should be enough to make stable diffusion circular. For qwen image, you need to also adjust the rope settings, see my PR here leejet/stable-diffusion.cpp#914 for the stable-diffusion-cpp side of it.

[CISC]

Please be aware you may have better luck submitting this PR (with tests) to llama.cpp instead, see CONTRIBUTING.md.