math: expose explicit engine namespaces cordic:: and dbl::

[NiceAndPeter]

Phase 3 of the math-engine plan — explicit engine namespaces

Surface both math engines by name so a single binary can mix them per call site, while the unqualified bnd::math::fn stays an alias for the build default.

Namespace	Engine	Availability
bnd::math::cordic::fn	integer / CORDIC	always (constexpr, FPU-free)
bnd::math::dbl::fn	double	unless BND_MATH_NO_FP
bnd::math::fn	the default	cordic under BND_MATH_FIXED/BND_MATH_NO_FP, else dbl

The qualified entry points carry the same signatures, domains, auto-deduced output grids, and domain static_asserts as the unqualified API — only the compute backend differs.

No new math, additive, low-risk

• cordic:: wraps the always-compiled *_impl cores; dbl:: wraps the *_core cores (plus the inline tan/signed-sqrt/pow_base/pow double logic).
• dbl:: is gated behind !BND_MATH_NO_FP: under the FP-free build a dbl:: call is a compile error; cordic:: always works.
• The existing public functions are untouched — zero regression risk; the namespaces are purely additive.

Proof

tests/test_math_engines.cpp instantiates both namespaces in one TU (templates only compile when used — the core proof), checks they agree on algebraically-exact inputs (sqrt(4), cos(0), …), that poles still error through expected under both, and that the unqualified name matches the build's default engine bit-for-bit.

Verification

• default 401/401, CORDIC 442/442
• all single-header smokes build (incl. single_header_nofp_smoke); amalgamation regenerated, drift check green

🤖 Generated with Claude Code