server: prevent data race from HTTP threads (#18263)

• server: prevent data race from HTTP threads

• fix params

• fix default_generation_settings

• nits: make handle_completions_impl looks less strange

• stricter const

• fix GGML_ASSERT(idx < states.size())

• move index to be managed by server_response_reader

• http: make sure req & res lifecycle are tied together

• fix compile

• fix index handling buggy

• fix data race for lora endpoint

• nits: fix shadow variable

• nits: revert redundant changes

• nits: correct naming for json_webui_settings

server: fix data race in to_json_anthropic (#18283)

release: update release workflow to store XCFramework as Zip file (#18284)

• Update release workflow to store XCFramework as Zip file

• Add comments to document Zip file requirement for XCFramework

• Apply suggestions from code review

Co-authored-by: Sigbjørn Skjæret sigbjorn.skjaeret@scala.com

Co-authored-by: Sigbjørn Skjæret sigbjorn.skjaeret@scala.com

tool/ex/tests: consistently free ctx, then model (#18168)

vulkan: Implement set_tensor_async and the event interfaces (#18047)

The goal is to enable the async loading code paths in
llama_model_loader::load_all_data, originally from #7896. This works and the
loads themselves are faster, but with host visible vidmem I think the cost of
allocating/mapping vidmem moves and becomes more expensive, and I don't see a
benefit by default. But with GGML_VK_DISABLE_HOST_VISIBLE_VIDMEM=1 I do see a
significant improvement in model loading time.

llama: fix RPC for -fit on (#18233)

vulkan: fix im2col overflowing maxworkgroupcount (#18180)

vulkan/cuda: fix topk_moe with exp_probs_b (#18071)

I updated test_topk_moe to more closely match llm_graph_context::build_moe_ffn
and added coverage for exp_probs_b and some other missing combinations. This
exposed a bug in both CUDA and Vulkan backends where they were assuming the
input to argsort and the input to get_rows are the same. I'd like to optimize
this graph in another change, but for now just get it functional.

CUDA also had a bug where it got n_experts from the wrong place, leading to
GGML_ASSERT failures in some of the new tests.

vulkan: support GGML_UNARY_OP_XIELU (#18062)

vulkan: in graph_optimize, try to group ADD operations (#18060)

I saw the adds not staying together in the new nemotron 3 nano model.