feat: sync internal release-0901 improvements

[lkygithub]

Description

Features

• Improve job placement and restart flows: preserve automatic versus manual placement when cloning or editing a Job, support scheduling nodes whose device resources are undiscovered or do not have a configured model without requesting fake resources, and account for resources released by the current Job during edit-and-restart.
• Improve Job lifecycle UX with consistent confirmation dialogs, clean restart of failed Jobs, completion notices, direct row actions, clearer phase derivation, compact Worker event details, readable long identifiers, and links from Worker cluster/node fields to their detail pages.
• Improve user and administrator login with RLark branding, accessible password visibility controls, and inline error feedback.
• Improve administrator node management with node overview statistics, managed node categories and metadata fields, safer batch metadata updates, and clearer resource information.
• Add cross-cluster network observability and resilience through Prometheus metrics, sidecar metrics/pprof serving, SSH application-level keepalives, reconnection tracking, and longer-lived idle connections.

Fixes and optimizations

• Make stop, restart, and delete cleanup wait for workloads and task PVCs, propagate restart annotations, record stopped end times, and allow Pending Jobs to transition correctly to terminal states.
• Surface abnormal Pod/container states more reliably by reconciling Pods through their owning Deployment, StatefulSet, or DaemonSet.
• Normalize image registry addresses and validate the behavior with unit tests.
• Run the rclone CSI controller with the privileges required for bidirectional mounts.
• Correct generated SSH jump commands and disable copying when the jump-host configuration is unavailable.
• Improve Worker layout, resource cards, node filtering, and responsive UI behavior.

Documentation and tests

• Update the English and Chinese Job, storage, and SSH key guides for the new lifecycle, placement, Worker detail, and SSH behaviors.
• Add coverage for Job state transitions and cleanup, restart annotation propagation, workload ownership, registry normalization, rclone manifest settings, login behavior, resource placement and availability, Job actions, and displayed phases.

Important

Task PVC lifecycle behavior changes in this release: stopping, restarting, or deleting a Job removes its task PVCs; starting or restarting creates empty task PVCs. hostPath data is preserved. The bilingual user documentation has been updated accordingly.

Motivation and Context

Source mapping:

• Previous synchronization baseline: internal release-0820
• Internal source tag: release-0901
• Internal source commit: 22bf783e26342e02df4eb1bb3d35ca6775beb9eb
• Public candidate commit: 1b6703317d1055a3084df54f7faa96f07c384f5e

How has this been tested?

The exact candidate commit was first validated in an isolated GitHub CI environment. All configured checks passed:

• Commit sign-off and PR title checks
• Go lint for api, apps/rlark, and apps/embodied-runtime
• Frontend ESLint and Prettier checks
• Strict MkDocs build and offline Markdown link validation
• Go unit tests for api, apps/rlark, and apps/embodied-runtime
• Docker builds for RLark, RLark UI, and embodied-runtime

Local validation also covered the Go test suites, frontend production build, frontend lint/format checks, and strict documentation build.

Types of changes

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Documentation update
• Breaking change (fix or feature that would cause existing functionality to not work as expected)

Checklist:

• My code follows the code style of this project.
• My change requires a change to the documentation.
• I have updated the documentation accordingly.
• I have added tests to cover my changes.
• All new and existing tests passed.