Fix dloglik crash when using CG optimizer with warped BLR models

[divye-joshi]

Fix dloglik crash when using CG optimizer with warped BLR models

Description

This PR fixes a bug where BLR.fit() crashed with a misleading Unknown function dloglik error when attempting to use the cg optimizer in combination with a warp function (e.g., warpsinharcsinh).

The root cause was that analytical gradients (dloglik) are not implemented for warped models. When the user manually selected optimizer="cg", fit() passed the unsupported dloglik function to the SciPy optimizer, which resulted in a hard crash during optimization.

Changes

• Added a validation check in BLR.fit() before optimization begins.
• If optimizer="cg" is used with a warped model, fit() will:
  • Log a warning indicating that analytical gradients are unsupported.
  • Automatically fall back to the default l-bfgs-b optimizer, which correctly uses numerical approximations.
• Added a new structured warning WARNINGS.BLR_CG_NOT_SUPPORTED_WITH_WARP in output.py for clarity.

Closes

Closes #378

[contsili]

very good catch! I only improved the error message inside the dloglik(). I don't expect this error to happen ever after the check you added but just to make sure i kept it and made it more clearly describing the issue