Bridges from the finite subgraph densities to the empirical-graphon formulas (#94, PR 1) #
The analytic half of the t/t_inj/t_ind interlude: on the equipartition step graphon of a
finite host H, the homomorphism density is SimpleGraph.t and the sampling mass is the
normalized exact-pullback count. Split from Graphon.SubgraphDensities so the pure density
API keeps a combinatorics-only import closure. PR 2 of #94 adds the collision-comparison
forms: the analytic homomorphism density is within k²/n of t_inj, and the sampling mass
within k²/n of t_ind — the quantitative content of "sampling ≈ injective sampling".
The homomorphism density of the empirical graphon is t: on the equipartition step
graphon of H, the analytic homomorphism density coincides with the combinatorial
density.
The sampling mass of the empirical graphon is the normalized exact-pullback count:
the probability that the k-sample of the empirical graphon of H equals G is the
proportion of all vertex maps pulling H back to exactly G.
The analytic collision comparison: the homomorphism density of the empirical graphon
is within k²/n of the injective density of the host.
The sampling collision comparison: the sampling mass of the empirical graphon is
within k²/n of the induced density of the host.