update torch, optuna & scikitlearn install deps

[hugolytics]

Hi All,

I'm encountering dependency conflicts when trying to install XGBoostLSS using Poetry. The package currently specifies strict maximum versions for several dependencies, including torch, optuna, and scikit-learn. This is causing conflicts with other common packages in my project that require more recent versions of these libraries.

Specifically:

1. torch: XGBoostLSS requires ~=2.1.2, but the current version is 2.3.0, which is needed by other packages in my project.
2. optuna: Required as ~=3.5.0, but newer versions might be necessary for compatibility with other libraries.
3. scikit-learn: Specified as ~=1.4.0, which may conflict with other packages requiring newer versions.

Would it be possible to update or relax these version requirements? This could involve:

1. Updating the dependencies to the latest stable versions that are compatible with XGBoostLSS.
2. Relaxing the version constraints to allow for a wider range of compatible versions (e.g., using ">=" instead of "~=" and specifying a higher maximum version).
3. If strict version requirements are necessary for certain dependencies, clearly documenting the reasons in the README or requirements file.

Here are some suggested edits for the setup.py file:

install_requires=[
    "xgboost>=2.0.0,<3.0.0",
    "torch>=2.1.0,<3.0.0",
    "pyro-ppl>=1.8.0,<2.0.0",
    "optuna>=3.0.0,<4.0.0",
    "scikit-learn>=1.0.0,<2.0.0",
    "numpy>=1.20.0,<2.0.0",
    "pandas>=1.0.0,<3.0.0",
    # ... (similar changes for other dependencies)
]