[CI/Build] Move test scripts and assets from .github/ to tests/ (part of #531)

[ighutake-debug]

Implements item 1 ("Move the scripts and assets from .github/ to `tests/") of the CI refactor plan agreed in #531 (comment).

Part of #531 — this PR covers only item 1 of 4; follow-up PRs will address the fake-server, job-consolidation, and Helm de-duplication items, so merging this should NOT close the issue.

What changed

Moved 15 loose CI files out of .github/:

• tests/scripts/: curl-02-two-pods.sh, curl-04-multiple-models.sh, curl-05-secure-vllm.sh, curl-06-monitoring.sh, port-forward.sh
• tests/assets/: values-01/04/05/06/07/08/09/10/11-*.yaml, template-chatml.jinja

Updated every referencing call site:

• .github/workflows/functionality-helm-chart.yml — helm install -f values paths + port-forward.sh invocations
• .github/workflows/router-e2e-test.yml — --chat-template path for the static-discovery backends
• tests/e2e/run-k8s-routing-test.sh — HELM_VALUES_FILE paths (10 references)
• tests/scripts/port-forward.sh — path to the curl scripts it dispatches

Also added tests/assets/** and tests/scripts/** to the paths filters of both workflows so changes to the moved files keep triggering CI.

No functional changes — pure move + reference updates (all 14 content files are 100% renames; only port-forward.sh has a 1-line path edit).

Validation done locally

• grep sweep: zero remaining references to the old .github/ paths
• bash -n passes on all moved/edited shell scripts
• YAML parse passes on both workflows and all moved values files
• pre-commit run on all changed files: actionlint, check-yaml, end-of-file, trailing-whitespace, codespell all pass

The self-hosted runner jobs (functionality-helm-chart.yml, router-e2e-test.yml) will exercise the moved scripts end-to-end on this PR's CI run.

BEFORE SUBMITTING, PLEASE READ THE CHECKLIST BELOW AND FILL IN THE DESCRIPTION ABOVE

---

• Make sure the code changes pass the pre-commit checks.
• Sign-off your commit by using -s when doing git commit
• Try to classify PRs for easy understanding of the type of changes, such as [Bugfix], [Feat], and [CI].

[gemini-code-assist]

Code Review

This pull request updates the paths in the end-to-end test and port-forwarding scripts to reflect the relocation of test assets and scripts from the .github/ directory to tests/assets/ and tests/scripts/. The review feedback suggests using $(dirname "$0") in port-forward.sh to make the script execution location-independent instead of hardcoding the relative path.

[gemini-code-assist]

Since port-forward.sh and the target scripts (e.g., curl-*.sh) are now located in the same directory (tests/scripts/), we can avoid hardcoding the relative path tests/scripts/ by using $(dirname "$0"). This makes the script location-independent and allows it to be run successfully from any working directory (including from within tests/scripts/ itself).

Suggested change

	bash "tests/scripts/$1.sh" "$ip" "$port"
	bash "$(dirname "$0")/$1.sh" "$ip" "$port"

[ighutake-debug]

Hey @ruizhang0101 — PR 1 is up as discussed in #531. Kept it strictly mechanical (pure moves + path updates; every content file is a 100% rename), so review should be quick. The self-hosted e2e jobs will exercise the moved scripts end-to-end on this run — I'll keep an eye on CI and fix anything that shakes out.

[ighutake-debug]

Follow-up for the record: the PR-time pre-commit.ci failure appears to have been a service-side runner error — on main, the repo's own pre-commit workflow passes with these changes included, and all four functionality-helm-chart jobs (which exercise the moved values files and port-forward.sh) are green. Nothing outstanding from this PR; @zerofishnoodles happy to adjust anything if you'd like changes in a follow-up.

[ruizhang0101]

LGTM

[ighutake-debug]

Status update and a correction:

1. Correction of my earlier note — I previously said the change was green on main; that run actually predated/excluded this change (main was later reset), so it did not validate these files. Apologies for the confusion.

2. Current CI on the reopened PR is almost fully green: pre-commit, shellcheck, checkov, helmlint, Python tests, CRD validation, both router discovery e2e suites, Secure-Minimal-Example, Minimal-Example-With-Monitoring, DCO, and the RTD build all pass.

3. The single failure (Two-Pods-Minimal-Example) does not look like a defect in this change. The step fails with helm install vllm ./helm -f tests/assets/values-01-2pods-minimal-example.yaml: no such file or directory. However, the sibling jobs in the same workflow run (Secure-Minimal-Example, Minimal-Example-With-Monitoring) opened their moved files at identical tests/assets/ paths successfully, so the files are present in the checkout — this looks like a stale workspace on the shared self-hosted runner for that one job.

@zerofishnoodles @ruizhang0101 — could you re-run just the failed Two-Pods job when convenient? Once it goes green, I'd appreciate a review whenever you have a moment. Thanks!