Feat/mlflow nested run support

[illorens]

Add nested run support to MLflow logger

---

Description

This PR extends MLflowLogger with support for nested MLflow runs, allowing child runs to be logged under a shared parent run. It also adds the ability to attach tags to both parent and child runs.

The logger now operates in two modes:

• Standard mode (parent_run_name=None): existing behaviour is preserved — reuses an active run if one exists, otherwise creates a new run named run_name.
• Nested mode (parent_run_name=<name>): creates or reuses a parent run named parent_run_name, then starts a nested child run named run_name. All metrics and parameters are logged to the child run.

New constructor parameters:

• parent_run_name — enables nested mode when set
• run_tags — tags applied to the child (or standard) run
• parent_tags — tags applied to the parent run in nested mode

No breaking changes in MLflowLogger. All existing call sites that omit the new parameters continue to work identically.

⚠️ Breaking change: The Trainer constructor parameters train_loss_fct and val_loss_fct have been renamed to train_loss_fn and val_loss_fn respectively. Any call sites passing these as keyword arguments will need to be updated.

---

Checklist

• Code is well documented and follows project conventions (linted, formatted)
• CHANGELOG updated (if applicable)
• License header is added to all new files
• For a new contributors, their name has been added to the AUTHORS file
• No sensitive data or secrets are exposed
• Unit tests added covering the new code

[hlasco]

There's no guarantee than run names are unique, I think you should run_id instead of run name