chore: simplify and encapsulate loggers

[floriankrb]

Description

This pull request refactors the logging setup in Anemoi Core to simplify and encapsulate logger creation for MLFlow, TensorBoard, and Weights & Biases.

Key Changes:
Logger setup functions (get_mlflow_logger, get_tensorboard_logger, get_wandb_logger) now take explicit parameters instead of relying directly on the full config, making them more modular and easier to test.
Minor cleanups and simplifications in handling paths, hyperparameters, and offline/resumed runs.

Impact:
Cleaner, more maintainable logging code.
Easier to extend with additional loggers or change configurations without touching the core training logic.
Reduces duplication and centralizes logger-related logic.

Related to #538

As a contributor to the Anemoi framework, please ensure that your changes include unit tests, updates to any affected dependencies and documentation, and have been tested in a parallel setting (i.e., with multiple GPUs). As a reviewer, you are also responsible for verifying these aspects and requesting changes if they are not adequately addressed. For guidelines about those please refer to https://anemoi.readthedocs.io/en/latest/

By opening this pull request, I affirm that all authors agree to the Contributor License Agreement.

[anaprietonem]

Nice @floriankrb! thanks for starting on this! Seems much clearer, do you plan to look at this further? wondering if this could be a good opportunity to discuss if we want support all those loggers and improve the unit tests for those