Assignment inference improvment

[tihsu99]

Inference previously assumed at most two resonances, leaving cases with R >= 3 incorrectly interpreted.

1. Assignment probability

   Shared resonance symmetry was previously ignored. For R resonances and daughter-symmetry order G:

   $$p(a) \sim \frac{1}{RG}$$

   Applying only the original G correction gives a peak at 1/R. Inference now uses:

   $$RG,p(a) \sim 1$$

2. Jet masking

   Padded jet indices (-1) previously participated in masking and could suppress valid assignments. These indices are now ignored.

3. Detection probability

   The previous recursive interpretation was only valid for R <= 2. It is replaced by:

   $$P(N \ge r) = \sum_{k=r}^{R_{\max}} P(N=k)$$

   This correctly supports any configured resonance multiplicity.