feat: add end-to-end priority queueing

[Kevin Cho (chokevin)]

Summary

• add TauGrid-managed default and priority tier selection across workload profiles, CLI rendering, and the Python SDK/CLI
• propagate the selected tier to Kueue Workloads and Job/RayJob pod templates, while rejecting ambiguous exact-class combinations
• expose resolved Kueue admission priority and Kubernetes pod priority in queue snapshots and the Portal overview
• order pending workloads by resolved numeric priority and then FIFO, with deterministic constrained-quota coverage
• add a distinct Active jobs section sourced from actual Job/RayJob runtime status so admission is not confused with execution
• lock down the portable Kueue ordering/preemption and PriorityClass contract with Helm tests and documentation

This is stack layer 2 above #323 and intentionally targets chokevin-portal-overview.

Validation

• cd portal && make test && make build && make lint
• cd portal/frontend && npm test -- --run — 36 tests passed
• focused core priority tests and go vet
• focused CLI/profile/Job/RayJob tests, go vet, and make build
• cd sdk/python/python && python3 -m pytest -q && ruff check . — 109 tests passed
• helm unittest charts/taugrid — 48 tests passed
• helm lint charts/taugrid
• helm template test charts/taugrid --namespace taugrid-system
• cd site && make check
• python3 scripts/check-license-headers.py
• git diff --check

No live cluster was contacted or modified.

AI assistance

GitHub Copilot assisted with implementation, test generation, conflict resolution, and validation. The resulting behavior and changes were reviewed and validated in the commands above.

[Pengfei Ni (feiskyer)]

LGTM