Polynomial long division carries its divisor's domain (#1174)

[Rafael-SOWNet]

n / d = quotient + remainder is only the quotient where d has a value. ln(x) / ln(x) divides out
to 1 + 0 / ln(x), which is 1 at x = 0 while the quotient it came from is undefined there —
ln(0) is a pole, and the remainder term carries only ln(x) != 0.

"ln(x) / ln(x)".Simplify()   was  1 provided not ln(x) = 0                 at x = 0: 1
                             is   1 provided not ln(x) = 0 and not x = 0   at x = 0: NaN

Closes #1174.

Why #1176 did not close it

#1176 corrected all seven cancellation-rule sites, and afterwards every registered set answered
correctly when asked directly:

Common / Power / CommonDenominator (+2 variants)  ->  1 provided not ln(x) = 0 and not x = 0
PolynomialLongDivision                            ->  0 + 1 + 0 / ln(x)

The public API still said 1, because this rewrite produced a shorter candidate and Simplify rated
it best. A rule being right is not the same as the answer being right, and only asking the entry
point tells them apart — which is why that PR said the issue stayed open rather than closing it.

Ordinary divisions are untouched

That is the whole reason this is safe: a divisor that cannot be undefined has a domain condition of
True, and it folds away.

(x^2 - 1) / (x - 1)        ->  x + 1 provided not x - 1 = 0        unchanged
(x^3 + 1) / (x + 1)        ->  x ^ 2 - x + 1 provided not x + 1 = 0  unchanged
(x^2 + 2*x + 1) / (x + 1)  ->  1 + x provided not 1 + x = 0        unchanged
x^2 / x, x^3 / x^2         ->  x provided not x = 0                unchanged

One recorded test moves

x!/x! now carries the factorial's domain instead of not x! = 0 — the same change
(x + 1)!/(x + 1)! took in #1176. That one went through a cancellation rule and this one through this
rewrite, which is why they moved a commit apart; the note in SimplifyTest that called the
difference out is updated rather than left describing a state that lasted one commit.

Measured at x = -4, -3, -2, -1, 0, 1, 3, 1/2 and -1/2: the old condition and the new one agree with
the unsimplified quotient at every point — NaN at the four gamma poles and 1 elsewhere. So the
printed form moves and the value does not. Recorded in BREAKING-CHANGES.md.

Both spellings changed together, MatchedRules and Patterns.cs.

Checks

Full suite 9560 passed, 0 failed, 14 skipped.

🤖 Generated with Claude Code

https://claude.ai/code/session_012sonx8iAspMiwRwokT1Ura