Update MoE training in example

[chenyushuo]

Description

As the title says.

Checklist

Please check the following items before code is ready to be reviewed.

• Code has passed all tests
• Docstrings have been added/updated in Google Style
• Documentation has been updated
• Code is ready for review

[Copilot]

Pull Request Overview

This PR updates the Megatron training configuration to add support for distributed checkpointing options, particularly for Mixture-of-Experts (MoE) models. The changes provide better guidance and configuration options for training MoE models with different checkpoint formats.

• Adds distributed checkpointing configuration parameters across all model components
• Updates documentation to clarify the two approaches for training MoE models
• Provides clearer instructions for users choosing between mBridge and manual model conversion

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
examples/ppo_countdown_megatron/train_countdown.yaml	Adds use_dist_checkpointing and dist_checkpointing_path parameters to actor, reference, and critic model configurations
docs/sphinx_doc/source/tutorial/example_megatron.md	Updates MoE training documentation with clearer instructions and adds the new checkpointing parameters to configuration examples

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

The documentation shows use_dist_checkpointing: true but the YAML configuration files consistently set this to false by default. This inconsistency could confuse users about the correct value to use when manually converting models.

Suggested change

	If you prefer not to use mBridge, set `use_mbridge: false`. Before training, you must first convert your Hugging Face model to the MCore format using the [Hugging Face to MCore converter](https://github.com/volcengine/verl/blob/main/scripts/converter_hf_to_mcore.py) from the **verl** repository. After conversion, update your config with:
	- `use_dist_checkpointing: true`
	- `dist_checkpointing_path: /PATH/TO/CONVERTED/MODEL/`
	If you prefer not to use mBridge, set `use_mbridge: false`. Before training, you must first convert your Hugging Face model to the MCore format using the [Hugging Face to MCore converter](https://github.com/volcengine/verl/blob/main/scripts/converter_hf_to_mcore.py) from the **verl** repository. After conversion, update your config as follows:
	```yaml
	megatron:
	use_mbridge: false
	use_dist_checkpointing: true
	dist_checkpointing_path: /PATH/TO/CONVERTED/MODEL/
	# ... other settings ...