ci: CI online test (#596)

* ci: add ci_test GitHub workflows

* fix: avoid cross-platform CUDA probing in tests

* ci: target master and setup Python in CI

* ci: use python module pip for CI dependency

* ci: update CI submodule for failure logs

* ci: update CI submodule for ci_ref scheduler

* ci: update CI submodule for source-mounted scheduler

* ci: update CI submodule for Unit generator fix

* ci: update CI submodule for tag model configs

* ci: update CI submodule for Unit failure logs

* ci: pass explicit devices for XPU unit jobs

* ci: standardize CI config extension to yml

* ci: update CI submodule for concise job names

* ci: update CI submodule for skipped job names

* Remove obsolete CI and lint config files

* ci: add manual platform dispatch

* ci: remove smoke and performance pipeline jobs

* Update Moore CI deployment fixes

* ci: rerun PR checks

* ci: default PR tests to nvidia

* ci: rerun nvidia check

* ci: update nvidia unit workflow

* ci: run PR checks on active platforms

* ci: register iluvatar platform

* ci: trigger checks on ci online branch

* ci: enable ascend online runner

* ci: rerun with metax scheduler fix

* ci: rerun metax after cancel

* ci: skip ascend image rebuild

* ci: rerun ascend with encoded args

* ci: rerun ascend after runner cleanup

* ci: rerun iluvatar with timeout guard

* ci: cancel stale online runs

* ci: cap metax unit runtime

* ci: match ascend runner label

* ci: avoid queued platforms blocking ascend

* ci: rerun ascend with runner proxy

* ci: rerun ascend after python compatibility fix

* ci: rerun ascend with scheduler image

* ci: rerun ascend locally

* ci: run metax quick operator subset

* ci: install ascend build dependencies

* ci: rerun iluvatar after scheduler fix

* ci: rerun metax quick subset

* ci: rerun with safe matrix output

* ci: rerun after matrix output fix

* ci: rerun after matrix output fix

* ci: rerun iluvatar after report fix

* ci: rerun ascend accepting docker 137

* ci: limit metax online smoke cases

* ci: rerun metax after busy gpu filter

* ci: rerun full ci online

* ci: address pr feedback

* ci: use prebuilt ascend test image

* test: generate fallback randint data on cpu

* test: format gemm skip reason as markdown

* ci: build ascend test image from dockerfile

* ci: update ci tooling submodule

* ci: opt ascend into buildkit

* ci: keep default repo branch on master

* ci: run ascend tests on free npu

* ci: let ascend pick an available npu

* ci: update dynamic ascend allocation tooling

* ci: update ascend npu allocation parser

* ci: update ascend logical device mapping

* ci: use nvidia base compatible with runner

* ci: align nvidia test command with master

* ci: run nvidia tests on compatible base image

* ci: address review comments

* ci: update moore resource locking

* ci: update scheduler stale lock cleanup

* ci: update nvidia gpu allocation

* ci: add v2 shadow workflow

* ci: handle unavailable v2 shadow agents

* ci: add v2 agent installer

* ci: match v2 runner labels

* ci: enforce v2 shadow checks

* ci: update v2 runner user agent

* ci: default v2 shadow to active platforms

* ci: limit v2 agent queue wait to ten minutes

* ci: use self-healing v2 agent workflow

* ci: use transient state dir fallback

* ci: use platform lock probe workflow

* ci: use checkout-free self-hosted workflow

* ci: use nested junit result detection

* ci: use per-job checked-out agent

* ci: use metax resource allocation fix

* test: keep tests aligned with master

* ci: update iluvatar ci tooling

* ci: use early-exit v2 queue watchdog

* ci: handle queued runners and update platform sets

* ci: add iluvatar runner filesystem repair script

* ci: enable iluvatar in legacy workflow

* ci: skip host gpu probing for iluvatar

* ci: pin iluvatar local runner support

* ci: pin shadow workflow ci ref

* Fix Iluvatar CI container setup

* Include Iluvatar CI build backend dependency

* ci: remove local iluvatar repair script

* ci: preflight runner availability before jobs

* ci: pin runner preflight token fix

* ci: pin best-effort runner preflight

---------

Co-authored-by: zhangyue207 <zhangyue207@users.noreply.github.com>
Co-authored-by: Vincent777 <140055255+Vincent777@users.noreply.github.com>
Co-authored-by: zkjh <zkjh@localhost.localdomain>

Author: 3 people