Sampling bound dev

[Kevnroux]

Summary

• Add optional manual Monte Carlo radial sampling bounds for all dwarf profiles in mixed-dwarf generation (useManualSamplingBounds, bound1/bound2 in for_developers.lua, wired through NBodyCtx.samplingBounds into mixeddwarf).
• When bounds are left at 0.0, keep profile-specific defaults; when set manually, recalculate each component’s mass as the enclosed mass within that bound.
• Fix and enable the Einasto dwarf profile (density/potential, set_model_params for h/d, 99.9% mass sampling bound, enclosed-mass support) and document it in the README.
• Fix first/second derivatives near the origin by switching to a forward stencil when r < 2h, so evaluations never probe negative radii.
• Extend recalculate_comp_mass beyond NFW/Cored to Plummer, Hernquist, King, and Einasto; rename shared cutoff gamma storage to const_gamma_func.
• Add plummer_einasto.lua to mixeddwarf_test, match Einasto’s KL radial window to the sampler, and fix failure counting (failed++ / test(s)).

Details

Manual sampling bounds

• New Lua toggles in for_developers.lua:
  • useManualSamplingBounds
  • bound1 / bound2 (kpc; 0.0 means “use code defaults”)
• Bounds are passed via NBodyCtx.samplingBounds and into predefinedModels.mixeddwarf{ samplingBounds = ... }.
• Defaults:
  • Plummer / Hernquist: 50 * (rscale_l + rscale_d)
  • NFW / Cored: rcut + 15*rdecay if cutoff set, else 5*r200
  • King: tidal radius
  • Einasto: radius enclosing ~99.9% of the mass using the Wilson-Hilferty approximation
• Manual bounds trigger enclosed-mass recalculation so particle masses stay consistent with the truncated sampling region.

Einasto

• Replace the old/broken unused formulas with Retana-Montenegro et al. (2012)–style density and potential.
• Store Einasto auxiliaries (n, h, d, const_gamma_func) on Dwarf.
• Remove the “do not use” warning; call set_model_params if needed.
• Add mixed-dwarf test coverage (plummer_einasto.lua) and KL bound helper for Einasto.

Numerical / mass recalculation

• Update the near-origin first/second derivative stencils to a forward formula when r < 2h. This was needed for Einasto, which blows up if the centered stencil probes negative radii.
• Enclosed mass can now be recomputed for additional profiles when the sampling bound changes (King via numerical dΨ/dr).