Fix plot_fit inconsistency with multi-component models

[zblz]

Summary

• plot_fit now sets a default e_range when modelfn is available on the sampler, ensuring all model curves (ML line, confidence bands, samples) are evaluated over a dense logspace energy grid
• Previously, the ML model line was extracted from emcee blobs evaluated only at the sparse data energy points, while confidence bands and samples were re-evaluated over a dense grid — this caused visual inconsistency, especially for multi-component models (e.g. SSC)
• _calc_ML, _calc_CI, plot_samples, and _plot_MLmodel all now receive the same e_range, producing smooth, consistent curves
• Updated docstrings for e_range and e_npoints parameters in plot_fit and plot_samples

Test plan

• Added test_calc_ML_uses_dense_energy_grid — verifies that _calc_ML evaluates the model over a 100-point dense grid and matches direct model evaluation
• Full test suite passes (tox -e py313 -- tests/)

Closes #240