feat(stdlib): STEP 4-B randomness + perf.now bindings (Refs #239, closes standards#327)

[hyperpolymath]

Adds the randomness + high-res-timer surface that estate property tests and benchmark fixtures need:

• bofig/tests/property/graph_properties_test.ts (377L) substitutes \${prefix}_${Math.floor(Math.random() * 1000000)}`` for ID generation.
• Bench tests want sub-millisecond timings (performance.now()).

What lands

stdlib/Deno.affine (+4 externs)

extern	lowers to	notes
math_random() -> Float	Math.random()	[0, 1). JS PRNG, non-crypto.
random_u32() -> Int	((Math.random() * 4294967296) >>> 0)	uniform u32
random_in_range(lo, hi) -> Int	Math.floor(Math.random() * (hi - lo)) + lo	uniform [lo, hi)
performance_now() -> Float	performance.now()	high-res sub-ms timer

math_random is the JS PRNG — not cryptographically secure. Sufficient for property-test input generation, sampling, and simulations. For cryptographic random bytes a separate crypto_random_bytes binding routing to crypto.getRandomValues() belongs in a different sub-issue (different host call, different threat model).

Tests

tests/codegen-deno/random_smoke.{affine,deno.js,harness.mjs} covers:

• 1000 math_random draws all in [0, 1)
• 10000 random_u32 draws all in [0, 2^32) with ≥ 1000 distinct values (catches degenerate-PRNG regression)
• 1000 random_in_range(0, 100) draws all in [0, 100)
• 500 random_in_range(50, 60) draws cover most of the window
• performance_now monotone non-decreasing across three consecutive calls (catches clock-resolution regression)

Verification

step	result
dune build bin/main.exe	✅
dune runtest	✅ 353/353
./tools/run_codegen_deno_tests.sh	✅ all harnesses (incl. the new one)

Out of scope

• Seeded RNG (@std/random) — defer until a determinism-needing property test surfaces.
• sleep / setTimeout — separate effect surface; not blocking STEP 4.

Relation to #504 + #507

The three STEP 3 / STEP 4-A / STEP 4-B PRs all add disjoint externs to stdlib/Deno.affine and lib/codegen_deno.ml. Each is independently mergeable from origin/main; expected merge order is #504 → #507 → this PR but the file sections are separate enough that the rebases are mechanical.

Refs

• Closes standards#327
• Refs: standards#239 (umbrella), standards#243 (STEP 4 per-repo unblock target), affinescript#504 (STEP 3 sibling), affinescript#507 (STEP 4-A sibling)

🤖 Generated with Claude Code