feat: support for SDXS-512 model (#1180)

* feat: add U-Net specials of SDXS

* docs: update distilled_sd.md for SDXS-512

* feat: for SDXS use AutoencoderTiny as the primary VAE

* docs: update distilled_sd.md for SDXS-512

* fix: SDXS code cleaning after review by stduhpf

* format code

* fix sdxs with --taesd-preview-only

---------

Co-authored-by: leejet <leejet714@gmail.com>

feat: add sd-server API support for steps, sampler and scheduler (#1173)

refactor: replace ggml_ext_attention with ggml_ext_attention_ext (#1185)

feat: enable vae tiling for vid gen (#1152)

* enable vae tiling for vid gen

* format code

* eliminate compilation warning

---------

Co-authored-by: leejet <leejet714@gmail.com>

ci: only push Docker images on master or release

fix(server): use has_file for mask multipart detection (#1178)

feat: prioritize gguf and safetensors formats for embeddings and LoRA…

chore: eliminate compilation warnings under MSVC (#1170)

chore: use Release build for windows-latest-cmake

fix: avoid issues when sigma_min is close to 0 (#1138)