[release test] Upgrade Ray Train tests on g3 machine to g4dn

[liulehui]

Why are these changes needed?

1. g3 instance is relatively old and have low availability on AWS.
2. upgrade to g4dn machine that have better availability to reduce test flakiness.
3. one successful release test for the golden_notebook_torch_tune_serve_test.aws

Related issue number

Checks

• I've signed off every commit(by using the -s flag, i.e., git commit -s) in this PR.
• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
  • I've added any new APIs to the API Reference. For example, if I added a
    method in Tune, I've added it in doc/source/tune/api/ under the
    corresponding .rst file.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[gemini-code-assist]

Code Review

This PR upgrades AWS instance types for release tests from g3 to g4dn to improve stability, which is a good initiative. Most changes are consistent with this goal. However, one change in release/golden_notebook_tests/gpu_tpl_aws.yaml introduces a g5g (ARM-based) instance, creating a mixed-architecture cluster that could cause compatibility issues. I've added a comment with a suggestion to align this with the other changes.

[liulehui]

one sample successful release test for the golden_notebook_torch_tune_serve_test.aws