feat: support multi-node TP/EP/DP/PP from training side, and TP/EP/DP from rollout side, with rdma, for models of deepseek arch#14

JensenFire merged 5 commits intoJD-ETH:jd/rdma-integrationJD-ETH/slime:jd/rdma-integrationfrom

JensenFire:jsf/patching_for_dskJensenFire/slime:jsf/patching_for_dskCopy head branch name to clipboard