2291cc5c

By: ValentinHirschi <valentin.hirschi@gmail.com>

add certified two-normal sampling component

Implement the graph-independent shared-energy pair chart with a full-circle
inverse, a certified residual disk and the existing normalized ordinary
fallback. Use the compiled eager active Jacobian and evaluate inverse density
at the supplied momentum point. Directed MPFR endpoint enclosures preserve
native values, including separated Quad limbs.

Migrate the existing component inverse/support contract and partition: separate
restricted coverage from prerequisite context, propagate certified exclusion
as zero support, gate proxies by actual support and require a selected full
sibling. Keep numerical uncertainty distinct from support exclusion. Reuse one
Esurface cut-root entry for physical and sampling callers with unchanged root
policy and a shared exact massless-origin derivative convention.

This is a represented-geometry component milestone. Graph/run-card binding,
shared host/CT preparation and prerequisite-only proposal decisions across
native rescues remain separate production gates. Direct-H still does not bound
the GL638 corner, and these tests make no new GL638 efficiency claim.

Validation:
- 161 focused core map/partition/native-root/raised-cut checks passed.
- All 9 joint/enclosure tests passed again after routine fixture initialization:
  cargo nextest run -p gammalooprs --lib --profile test_gammaloop --retries 0
    -E 'test(sampling_joint::) | test(joint_bridge) | test(joint_density_partition) | test(native_mpfr_enclosures)'
- Saved-state reference acceptance and summed/canonical-channel API checks passed:
  cargo nextest run -p gammaloop-api --lib --retries 0
    -E 'test(saved_generated_process_runs_reference_acceptance_after_reload) | test(summed_sampling_matches_explicit_canonical_channels)'
- Core/API Python-feature cargo checks, formatting and diff check passed.
- Core all-target clippy has no warnings on changed lines; existing warnings remain.

The real component tests cover unequal/massless energies, native perturbations,
both circle signs, an independent determinant, complete Gaussian/raw moments,
foreign conditional affine support and an analytic finite J*w/R limit from
either selected channel. Existing gammaloop_state storage layout is unchanged;
no new runtime card syntax is enabled by this component alone.

User prompt:
Just a status question—please continue your current orchestration unchanged: where does the joint two-normal density sit in the implementation sequence, and what prerequisites remain before you reach it? (direct-H focusing alone is not expected to bound the previously identified corner’s weights.)

Timeline