A power of x past the twelfth beside a block is split off

[Rafael-SOWNet]

Part of #718.

N/(x^k B) is split as P/x^k + R/B, where P is the first k terms of the series of N/B at 0. The route that does this capped k at 12. Rubi's 1.1.3.8 goes up to x^17 beside (a + b x^3)^3. The series costs one term per power of x, so this branch gives k its own cap, MaximumPowerOfXOverABlock = 24. The block's degree, and the other routes that shared the old constant, keep 12.

Seven rows from 1.1.3.8 now answer. 2.5.0 and master 3c86dda7 both declined them:

integrand	length	time
(c + d x^3 + g x^6 + f x^9)/(x^14 (a + b x^3))	805	0.4 s
(c + d x^3 + g x^6 + f x^9)/(x^15 (a + b x^3))	823	0.05 s
(c + d x^3 + g x^6 + f x^9)/(x^17 (a + b x^3))	903	0.05 s
(c + d x^3 + g x^6 + f x^9)/(x^13 (a + b x^3)^2)	793	7.4 s
(c + d x^3 + g x^6 + f x^9)/(x^14 (a + b x^3)^2)	1,308	0.15 s
(c + d x^3 + g x^6 + f x^9)/(x^13 (a + b x^3)^3)	1,172	4.1 s
(c + d x^3 + g x^6 + f x^9)/(x^14 (a + b x^3)^3)	1,931	0.3 s

Lengths are characters of Stringize(). All seven ran in one process, in this order.

Rubi corpus, both arms built from 88354428, fix against master:

• Files 0–8, plus a 14,154-row pocket: the pocket goes from 13,969 to 13,976 solved. Every file has the same counts in both arms, and both arms have 0 wrong.
• Rows that moved: the seven above, from unevaluated to solved. A rerun of only those rows gives 7/7 solved on this branch and 0/7 on master, all graded by differentiating back.
• Allocation gate: passed.
• Unit suite: 15,196 passed, 0 failed, 13 skipped (net10.0).

BREAKING-CHANGES.md has an entry, A power of x past the twelfth beside a block is split off. AHighPowerOfXBesideABlockIntegralTest integrates the x^14 and x^17 rows over a + b x^3, and the x^14 row over its square, and checks that each answer differentiates back.

🤖 Generated with Claude Code

https://claude.ai/code/session_012sonx8iAspMiwRwokT1Ura