Consolidate PEFT support into Megatron-LM

## Goal

Move the PEFT support required by Megatron-Bridge into Megatron-LM so Bridge PEFT recipes can use `megatron.training` as the canonical training loop.

## Scope

- Move performant LoRA, canonical LoRA, DoRA linear adapters, and the `LoRALinear` wrapper into Megatron Core.
- Apply adapters and freeze the base model before model wrapping in the Megatron-LM training loop.
- Add adapter-only checkpoint save and load support to `megatron.training.checkpointing`.
- Migrate Megatron-Bridge PEFT recipes to the Megatron-LM training loop.

## Completion criteria

- Existing Megatron-Bridge PEFT APIs continue to work through the migration.
- Existing PEFT recipes retain their current behavior when run through the Megatron-LM training loop.
- Representative performant LoRA, canonical LoRA, and DoRA recipes preserve convergence and performance within agreed tolerances.
- Adapter-only checkpointing, adapter merging, and adapter export continue to work.

Start by reading the Megatron-LM training loop and megatron.training.checkpointing, then compare the existing Megatron-Bridge PEFT APIs and recipes. Done means the PEFT implementations and migration preserve recipe behavior, adapter-only checkpointing, merging, export, convergence, and performance within agreed tolerances.

Written by the indexing model from the issue text.