noun: math.hoon transcendental jets (@rd/@rs/@rh/@rq)

[sigilante]

Registers the math.hoon transcendental library jets (from the numerics
project's /lib/math) in the userspace %non chapter, so the four IEEE
precision doors run native SoftFloat instead of interpreted Hoon.

What

• 15 transcendentals × 4 doors — exp log sin cos tan atan atan2 asin acos sqt cbt pow pow-n log-2 log-10 for @rd (f64), @rs (f32), @rh (f16,
  native), @rq (f128) — registered as non → math → <door> → arm in the
  135, 136 and 137 trees.
• Jet bodies in jets/i/math.c (chub-based I/O, word-size-agnostic), the
  u3qi_*/u3wi_* decls in q.h/w.h, and the build.zig source entry.

Verification

On a fresh hoon-135 fakezod: all four doors fire (exp:rd/rs/rh:math
~1 µs/call, exp:rq ~1.8 µs, vs ~250 µs interpreted) and -test %/tests/lib
is 247/247 green, bit-exact to the Hoon door (same reduction, coefficients,
Horner order). Jet ≡ interpreted Hoon confirmed arm-by-arm.

🤖 Generated with Claude Code

[sigilante]

CI failure: near is a reserved identifier/macro on the CI toolchain (it compiles on macOS clang but not Linux gcc — near is a legacy reserved word). That hits the @rh inverse-trig code in math.c and affects both vere PRs.

[sigilante]

Dependency note: this PR is based directly on develop and is independent of #1021 (the 32-bit lagoon→SoftBLAS migration) — no shared files, no dependency, so the two can merge in either order. Its 64-bit counterpart #1045 is stacked on #1022.

[sigilante]

Rebased onto post-vere64 develop (single-source tree now builds both loom widths). Since the whole math.c is new on develop, only the shared jet-registration files needed merging; math.c is unchanged in substance from the prior head plus two rebase adaptations, folded into the last commit:

• The four single-argument @r jet helpers type their u3r_at(u3x_sam, cor) sample u3_weak before the u3_none/u3ud guard (refcount linter, refcount linter #1059).
• The five u3r_mean call sites move from the old flat, 0-terminated form to develop's {axe, &out} braced-pair form (u3r_vmean), which -Werror/-Wmissing-braces now requires.

Verification: builds clean at both widths (zig build, zig build -Dvere64, -Werror); refcount linter reports 0 findings on math.c (strict-weak). The numerics libmath math test suite (12 files: exp/log/sqrt/tan/trig/atan/ainv/rh/rq/derived/rounding/constants) runs green jetted on a fresh fakezod for both binaries (272 OK arms 32-bit, 411 64-bit, 0 failures).

Note: numerics#74 (the _math_rnd file-static leaving u3qi_* composites call-history-dependent) is a latent API-hygiene issue — no in-tree caller reaches a composite without its u3wi_* wrapper setting the mode first — and the clean fix is a rounding-mode-parameter refactor across ~12 entry points. Left as a separate follow-up rather than bundled into this rebase.

This makes #1045 (the separate 64-bit math port stacked on #1022) redundant; it can be closed once this lands.