[examples] Add dynamic context parallel example

[ilml]

Summary

• Add a minimal long-context example for Megatron-Core dynamic context parallelism (hybrid_context_parallel).
• Show the Bridge config knobs, HybridCPDataLoaderWrapper, and a custom forward step that consumes per-sample local_cp_size.
• Include a --dry-run scheduler preview and a tiny packed toy dataset for a 2-GPU smoke run.

Testing

• uv run python -m py_compile examples/long_context/dynamic_context_parallel.py
• RUFF_CACHE_DIR=/tmp/megatron_bridge_ruff_cache uv run ruff check examples/long_context/dynamic_context_parallel.py
• RUFF_CACHE_DIR=/tmp/megatron_bridge_ruff_cache uv run ruff format --check examples/long_context/dynamic_context_parallel.py
• RUFF_CACHE_DIR=/tmp/megatron_bridge_ruff_cache uv run pre-commit run --all-files
• tmux window 0:2, Docker container 7af67331b54b: uv run python -m torch.distributed.run --master_port=29513 --nproc_per_node=2 examples/long_context/dynamic_context_parallel.py --eval-iters 1 (__EXIT_CODE:0__)

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.