Enhancement: Output per-pixel areal density maps from SAMMY fitting

[KedoKudo]

Context

The imaging pipeline currently outputs normalized atom-fraction composition maps (summing to 1.0 per pixel) via ResultsAggregator. This is correct for composition analysis.

However, the absolute areal density (atoms/barn-cm) per isotope per pixel is also scientifically valuable — it encodes both composition and sample thickness variation across the image.

Current behavior

• Pass 1 fits the global thickness n(x,y) per pixel
• Pass 2 fits per-isotope abundances f_i(x,y) (IFLISO=1)
• The LPT parser (lpt_manager.py) extracts only abundances and chi-squared
• ResultsAggregator normalizes abundances: w_i = f_i / Σf_i
• Per-pixel thickness n(x,y) is discarded after Pass 1

Desired behavior

In addition to normalized composition maps, optionally output:

1. Per-pixel thickness map: n(x,y) in atoms/barn-cm — extract from Pass 1 LPT or PAR output
2. Areal density maps: n(x,y) * f_i(x,y) for each isotope — the product of thickness and abundance
3. Store both in Imaging2DResults and HDF5 output

Physics

T(E) = exp( -n * Σᵢ fᵢ * σᵢ(E) )

areal_density_A0(x,y) = n(x,y) * f_A0(x,y)    [atoms/barn-cm]
composition_A0(x,y)   = f_A0 / Σᵢ fᵢ            [atom fraction, sums to 1.0]

Implementation notes

• Extract fitted thickness from SAMMY LPT (broadening parameters section) or from the output PAR file after Pass 1
• Add thickness_map: Optional[np.ndarray] and areal_density_maps: Optional[np.ndarray] to Imaging2DResults
• Backwards compatible: existing code that only uses abundance_maps is unaffected

Priority

Low — composition maps are sufficient for current research. This is a nice-to-have for quantitative analysis.