ci: server: fix windows is not building PR branch

ci: server: fix server tests after ggml-org#6638

ci: add ubuntu latest release and fix missing build number (mac & ubu…

…ntu) (ggml-org#6748)

ci: bench: artefact name perf job

ci: server: debug

server: bench: fix graph, fix output artifact

server: bench: init

TMP - bench workflow

TMP - bench workflow

common: llama_load_model_from_url using --model-url (ggml-org#6098)

* common: llama_load_model_from_url with libcurl dependency

Co-authored-by: Georgi Gerganov <[email protected]>