Read csch as the difference of the exponentials with its half gathered elsewhere

[Rafael-SOWNet]

csch(y) arrives as (e^y - e^-y)^(-1) beside a 2 the product gathered. ReadTheHyperbolicFunctions recognised the difference only with its half, (e^y - e^-y)/2. It now reads e^y - e^-y as 2 sinh(y) and e^y + e^-y as 2 cosh(y), only as the base of a whole power. csch(c + d x) (a + b sinh(c + d x)^4)^n ran past fifteen minutes on 2.5.0 and is answered in under half a second.

Part of #718 — Rubi's 6.1.7:237 and :247.

The whole-power restriction is measured. Read everywhere, the bare difference lost 27 rows of family 6 that master answers (x/csch(x)^(3/2) + …, sqrt(a + i a csch(c + d x)), …). There the reading handed a root to a rule that then declined it. Restricted, all 27 are answered again.

Measured against master b2ef509:

• family 6 whole (3,690 rows): the two gains, 0 losses. The 3 other moves (6.1.5:396, 6.4.2:48, 6.7.1:928) were re-run alone on both arms and agree; they are slow rows under load. 6.1.5:94 is WRONG on both arms, as before;
• the chain (family 1 at 40 a file, family 0 whole, families 2–8 sampled): no moves; the benchmark gate passed; the suite passed (15,551). After merging master 53a0fa4: built for every target, suite 15,563 passed.

Recorded in BREAKING-CHANGES.md.

🤖 Generated with Claude Code

https://claude.ai/code/session_012sonx8iAspMiwRwokT1Ura