Experimental: assay unused Gaussian/Fock and coherent-routing methods

[Inserian]

Bounded experimental lane only. This PR does not change QSA production/runtime behavior and remains draft/unmerged.

The first frontier gate is now positive across three previously unused method families:

• Gaussian bosonic / fixed-particle Fock structural representations. The analytic assay passes coherent displacement, single-mode squeezing, two-mode squeezing and purity checks at floating-point error. At 128 Gaussian modes with a 32-state local cutoff, moment storage is 526,336 bytes versus the corresponding explicit truncated Fock statevector, a structural storage ratio of about 1.3869e188 for the certified Gaussian family. The fixed-particle 256-mode/32-particle sector has 2.8341198877941199e42 legal configurations while one occupation tuple is 1,024 bytes. These are representation results only; arbitrary non-Gaussian states remain outside the certificate.
• Native Gaussian candidate. include/qubit/qgaussian.hpp and src/qgaussian.cpp now implement vacuum first/second moments, displacement, phase shift, squeezing, beam splitters, two-mode squeezing, photon-number observables and fail-closed validation. A standalone C++20 adversarial test covers analytic values, passive number conservation, large 256-mode state, invalid modes and nonfinite inputs. The native gate passed in workflow 31649085304. It is not yet wired into the production representation broker.
• Coherent route recombination. With identical route populations, 2,048 paired phase cases produced mean coherent output gap 0.1266103506 while the dephased control gap was about 1.0e-17. A 16-code orthogonal phase assay had coherent diagonal 1.0, off-diagonal <=3.75e-33, while the dephased output was the same 0.0625 for every code. This is causal phase-information evidence, not yet a Brain capability claim. Frozen Brain routing-trace replay remains the next Brain-facing gate.
• QTT Schrodinger-Poisson eligibility. The new dyadic-field assay passed from 256 through 65,536 grid points with rank budget 32. At 65,536 points, all tested fields remained at max TT rank 13 after a self-consistent Poisson solve and split-step evolution; norm drift was 1.78e-15 and relative Poisson residual 3.21e-14. The evolved complex field used 18,592 TT bytes versus 1,048,576 dense bytes (56.4x); density used 5,344 bytes (196.2x), and potential used 6,240 bytes (168.0x). This assay deliberately materializes the fields to certify QTT ranks, so it does not yet claim an end-to-end QTT-native speedup. The next field gate is a non-materializing native QTT Laplacian/Poisson/evolution path.

Decision-bearing workflow: QSA Frontier Unused Methods v1, run 31649085304, success. Evidence artifact ID 9161877369, uploaded ZIP SHA-256 a0178e8dd173728b6d3ea15944e3d937459c8291f49c7b27826584868fe9b421.

Base remains the accepted QSA PR62 head 54adffdf5cf05895d02974346d341b1e23d9c6c0. Do not merge this frontier PR until the native representation and matched-control gates are complete.