fix: training-integration uat by updating image

[orfeas-k]

As the updated notebook description mentions,

• Update PyTorchJob image according to upstream E2E tests. The only difference is that we use v1-3a360ba (the most recently pushed image) to avoid using latest, since this can result in inconsistent test runs.
• Update registry from which PaddleJob image is pulled to follow upstream E2E tests.

Closes canonical/bundle-kubeflow#894, canonical/bundle-kubeflow#910

Test the PR

1. Deploy CKF in any environment. It turns out that this issue was the same in Microk8s, AKS, EKS everywhere with juju 3.5.0 agent
2. Run training integration UAT either in a notebook or by checking out to this branch in uats repo and running tox -e kubeflow-remote -- --filter="training"
3. See it succeeding

[DnPlas]

Two tiny comments, other than that LGTM.

[DnPlas]

I can confirm the tests pass now:

============================= test session starts ==============================
platform linux -- Python 3.8.10, pytest-8.2.2, pluggy-1.5.0 -- /opt/conda/bin/python3.8
cachedir: .pytest_cache
rootdir: /tests/.worktrees/bd99a83476a34b1ea6c6a7d9ca4ceff6191eab3d/tests
configfile: pytest.ini
plugins: anyio-3.6.2
collecting ... collected 9 items / 8 deselected / 1 selected

test_notebooks.py::test_notebook[training-integration]
-------------------------------- live log call ---------------------------------
INFO     test_notebooks:test_notebooks.py:44 Running training-integration.ipynb...
PASSED                                                                   [100%]

================= 1 passed, 8 deselected in 917.94s (0:15:17) ==================
PASSED
------------------------------------------------------------------------------- live log teardown --------------------------------------------------------------------------------
INFO     test_kubeflow_workloads:test_kubeflow_workloads.py:82 Deleting Profile test-kubeflow...
INFO     httpx:_client.py:1013 HTTP Request: DELETE https://172.31.15.25:16443/apis/kubeflow.org/v1/profiles/test-kubeflow "HTTP/1.1 200 OK"
INFO     test_kubeflow_workloads:test_kubeflow_workloads.py:141 Deleting Job test-kubeflow/test-kubeflow...
INFO     httpx:_client.py:1013 HTTP Request: DELETE https://172.31.15.25:16443/apis/batch/v1/namespaces/test-kubeflow/jobs/test-kubeflow "HTTP/1.1 200 OK"


========================================================================= 2 passed in 1154.92s (0:19:14) =========================================================================
  kubeflow-remote: OK (1172.59=setup[16.52]+cmd[1156.07] seconds)
  congratulations :) (1172.67 seconds)