[aman-coder03]

What does this PR do?

This PR refactors XLM's output tracing to align with the standardized output capturing patterns used across the codebase.

Key changes:

• Refactors transformer blocks into a dedicated XLMLayer module to enable structured output capture
• Integrates the capture_outputs decorator with _can_record_outputs to automatically capture attention weights from each transformer layer
• Removes manual attention propagation logic and relies on the standardized output capturing infrastructure
• Implements explicit hidden_states collection in XLMModel.forward to ensure embedding outputs and all intermediate layer outputs are correctly returned
• Ensures proper handling and propagation of output_attentions, output_hidden_states, and return_dict flags

Why is this needed?

This aligns XLM with the standardized output tracing architecture introduced in #43979 ensuring consistent behavior across models and compatibility with the shared output capturing infrastructure.

[github-actions]

[For maintainers] Suggested jobs to run (before merge)

run-slow: flaubert, xlm

[ArthurZucker]

Thanks for the XLM refactor, @aman-coder03! The #43979 call to contributions was closed right after this was opened and we couldn't review all the PRs it produced, so I'm closing this one. Sorry it didn't get looked at in time.