📌 Updated Prior Passing Behaviour

This PR changes the default behaviour for how priors are passed from one model fit to the next.

Previously, priors were automatically updated to be centered on the median of the posterior from a previous result — using truncated Gaussians whose parameters reflected that inferred solution. While useful when samplers were less robust, this approach could introduce subtle shifts in results depending on prior updates.

Now, by default, priors are passed using the original priors defined in the configuration, without modification based on the previous fit.

✅ Benefits:

• Results are now fully reproducible and no longer depend on arbitrary prior updates between fits.
• The behaviour is more transparent and consistent, especially for workflows involving chaining models or rerunning fits.

🔄 Legacy Behaviour Still Available:

The previous behaviour — centering priors around the last result — is still available via the model_centred API, which explicitly builds new priors based on the previous fit’s inferred values.

This change improves reproducibility while still allowing flexible workflows when prior adaptation is needed.