feat(v4.0.0): Q.COS, bit for bit with v3

Q.COS is v3's q48_cos_approx: on |(Q.REDUCE) x|, the even Taylor series
to n = 10. It sets term and sum to 1.0, n to 2 and the sign cell to 0,
then jumps into Q.SIN's loop, so it adds no call level.

Checked bit for bit against v3's q48_cos_approx (ported into the test)
on the same 33 edge angles and 3000 pseudo-random angles as Q.SIN, at
32- and 64-bit cells, optimised and ASan+UBSan (`make test`,
`make sanitize`). Mutating the start index, the start sum, the
alternation or the sign cell each fails 2500+ cosine checks and no sine
checks.

Headroom (data under arg / return): Q.COS 3/2.

Test results on the amd64 host only. This is a development check, not
acceptance (JUSTIFICATION.md section 16).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

docs/v4.0.0/DECOMPOSITION.md

@@ -147,8 +147,8 @@ IF body1 ELSE body2 THEN →  if L1  drop body1  jump L2
 times, as on the F18. `FOR ... UNEXT` is the same but the body must fit in one instruction word.
 
 **Register conventions.** `A` and `B` are caller-saved. A word that uses them says so. Words in this
-document that clobber `A`: `@ ! +! -! 2@ 2! C@ C! UM* * UM/MOD Q.FROM-INT Q.TO-INT Q.* Q./ Q.EXP Q.SQRT Q.LOG Q.SIN SEND RECV`. Words that clobber `B`:
-`Q./ Q.EXP Q.SQRT Q.LOG Q.SIN SEND RECV`.
+document that clobber `A`: `@ ! +! -! 2@ 2! C@ C! UM* * UM/MOD Q.FROM-INT Q.TO-INT Q.* Q./ Q.EXP Q.SQRT Q.LOG Q.SIN Q.COS SEND RECV`. Words that clobber `B`:
+`Q./ Q.EXP Q.SQRT Q.LOG Q.SIN Q.COS SEND RECV`.
 
 **Return-stack words** (`>R R> R@ 2>R 2R> 2R@ I J UNLOOP` and the loop runtimes) are always IN.
 
@@ -664,7 +664,7 @@ word. Per D-10 it occupies two cells on a 64-bit node too. Per D-8, v4 Q values
 | `D2*C` | CAP | `( lo hi cin -- lo' hi' cout )`: the double shifted left one bit, `cin` entering at the bottom and `cout` the bit leaving the top (both 0 or 1). `a!  dup -if L1 drop 1 jump L2  L1: drop 0  L2: push  over -if L3 drop 1 jump L4  L3: drop 0  L4:  over + +  push 2* a + pop  pop` — `cin` waits in `A`, and `2hi + m` is `over + +`, so it needs one return entry. Executed on the golden model (2026-10-02). Clobbers `A`. |
 | `(Q/)` | CAP | Variable, 5 cells: `Q./`'s quotient register, divisor and result sign. Like `BASE` and the hold buffer, it lives in node memory. |
 | `Q.ABS` `Q.NEG` | CAP | `DABS`, `DNEGATE` — executed on the golden model (2026-10-02) against v3's `q48_abs` and `0 - q`: bit-for-bit at 32-bit cells, including v3's wrap of Q min to itself; at 64-bit cells the low cell is v3's and the high cell is the true sign, per D-10. |
-| `Q.COS` | CAP | Algorithm ported from `q48_words.c`; the hosted C version is the golden model. |
+| `Q.COS` | CAP | v3's `q48_cos_approx`: on `xu = |(Q.REDUCE) x|`, `1 - xu^2/2! + xu^4/4! - …` to `n = 10`, run by `Q.SIN`'s loop: `(Q.REDUCE)  -if A1 inv 1 + A1:  0 (QT) b! !b  0 over 0 Q.* drop (QT) 1 + b! !b  65536 (QT) 2 + b! !b  65536 (QT) 3 + b! !b  2 (QT) 4 + b! !b  -1 (QT) 5 + b! !b  jump TRIG` — term and sum start at 1.0, `n` at 2, and the sign cell is 0 so nothing is negated at the end. It jumps into the loop rather than calling it, so it costs no extra return entry. Executed on the golden model (2026-10-03): bit for bit v3's result on angles of every size and sign. Leaves its caller 3 data cells under its argument and 2 return entries. Clobbers `A` and `B`. |
 | `Q.SIN` | CAP | v3's `q48_sin_approx`: on `r = (Q.REDUCE) x` and `xu = |r|`, `xu - xu^3/3! + xu^5/5! - …` to `n = 11`, each term the last times `xu^2` over `n(n-1)`, stopping once a term is below 10 ulp; negated if `r < 0`. After the reduction every value fits one cell, so the variable `(QT)` holds single cells and the only calls are `(Q.REDUCE)`, `Q.*`, `Q./` and `UM*`. The Taylor loop is written so that `Q.COS` can jump into it. Executed on the golden model (2026-10-03): bit for bit v3's result on angles of every size and sign. Leaves its caller 3 data cells under its argument and 2 return entries. Clobbers `A` and `B`. |
 | `(Q.REDUCE)` | CAP | `( x -- r )`, v3's `q48_reduce_angle`: the angle reduced to one cell, `-pi <= r <= pi` with `pi = 205887`, `2pi = 411774` — the remainder of `x / 2pi` with the sign of `x`, then one step of `2pi` back into range. `x / 2pi` does not fit a cell, but only the remainder is wanted: `|x| mod 2pi` is two `UM/MOD` steps, the high cell first and its remainder then leading the low cell. `dup (QT) b! !b  -if L1 DNEGATE L1:  0 411774 UM/MOD drop  411774 UM/MOD drop  (QT) b! @b -if P1  drop inv 1 + jump J1  P1: drop  J1:  dup -205888 + -if BIG  drop  dup 205887 + -if OK  drop 411774 + ;  OK: drop ;  BIG: drop -411774 + ;` Executed on the golden model (2026-10-03), bit for bit v3's. Clobbers `A` and `B`. |
 | `(QT)` | CAP | Variable, 6 cells: `Q.SIN` / `Q.COS`'s sign, `x^2`, term, sum, `n` and subtract flag. |