[rozyczko]

Variable Resolution Support:

• Each dataset is now shown, fitted, and reported with its own measured resolution, rather than inheriting the resolution from the first dataset. This affects the Analysis page, report figures, Bayesian posterior bands, and residuals.
• Bayesian sampling now uses the correct resolution for each dataset by leveraging the library's Project.prepare_sample(), ensuring consistency between classical and Bayesian fits when multiple datasets are present.
• The model spin cross-sections are calculated without smearing by passing a zero width to the calculation, preventing interference with concurrent fits and ensuring correct visualizations.

Data Import and User Feedback:

• Added support for merging multiple measured curves (e.g., different angles or wavelength bands) into a single experiment, with each point retaining its own resolution. Files without resolution columns are added with an assumed resolution and a warning, and files that cannot be added are reported and skipped.
• Introduced a new loadWarning signal in both the Python and QML backends to notify users when an import succeeds with caveats (e.g., missing resolution information).

Codebase and API Adjustments:

• Refactored the Bayesian sampling preparation to remove the use of scipp DataGroup, relying instead on the library's own fitter preparation, which handles per-dataset resolution.
• Updated plotting and reporting logic to consistently use the correct resolution function for each dataset, ensuring that overlays and residuals match the data as fitted