Test fixes

[terrykong]

1. mcore generation config: fix: mcore generation config restored in nightly test #1720
2. gemma skip tokenizer fix: fix: gemma3 27b must now have skip_tokenizer_init=False in vllm #1721
3. seq parallel + tp no longer crashing: fix: remove seq_parallel + tp restriction in dtensor v2 #1725
4. cpu offload bug in v1: fix: apply offloading change from v2 to v1 #1726
5. rm checkpoint dir if successful (same PR as 5)
6. median metric change: fix: use median instead of mean for logprob error for stability in nightlies #1722
7. logger fix (val metrics were skipped) fix: log metrics that can be coerced to scalars #1723
8. increase time for some tests that failed due to model download fix: fix several nightly tests that were flaky #1724

[github-actions]

ℹ️ File Consistency Check

Check based on commit: 4d435d8 (PR #1718 from test-fixes)

✅ DTensor Policy Worker Synchronization Check

Both DTensor policy worker files were modified in this PR:

• nemo_rl/models/policy/workers/dtensor_policy_worker.py
• nemo_rl/models/policy/workers/dtensor_policy_worker_v2.py

Please ensure that the changes are consistent between both files where applicable.

---

_{This check ensures that related file implementations remain synchronized across the codebase. If you believe this warning is incorrect or the files should intentionally differ, please add a comment explaining the reasoning.}