Say what the three-argument integral actually cost, measured on 1.3.0 (#875)

The 1.4.0 entry showed one row, integral(f, x, 1), which 1.3.0 read as integral(f, x). That
makes the removal look like a no-op losing a no-op. It was not: a count of two or more was a
real iterated integral, and it is the part that is gone.

    1.3.0   integral(x, x, 2)             -> x ^ 3 / 6
    1.4.0   integral(x, x, 2)             -> FunctionArgumentCountException

The entry already said "for a repeated antiderivative, nest the calls", which is right and is
now the row that shows why it is needed. But following that advice does not give back what was
there, because #657 also began adding the constants of integration:

    1.3.0   integral(x, x, 2)             -> x ^ 3 / 6
    1.3.0   integral(integral(x, x), x)   -> x ^ 3 / 6
    1.4.0   integral(integral(x, x), x)   -> C_1 + C * x + x ^ 3 / 6
    master  integral(integral(x, x), x)   -> C_1 + C * x + x ^ 3 / 6

So the migration path is right but not answer-for-answer, and someone who takes it and diffs
the output should be told which difference to expect rather than discovering it.

The last row of the table also read "--" for integral(f, x, a, b) on 1.3.0, where the measured
value is a FunctionArgumentCountException naming three arguments or two -- the mirror image of
the message 1.4.0 gives for three. This file asks for the value on the previous release rather
than a dash, so it now gives it.

Each line above measured by parsing the string in a process referencing that published package,
1.3.0 and 1.4.0 from nuget and master built from source. Documentation only; no code changed.

Closes #847.

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>

Author: Rafael-SOWNet

BREAKING-CHANGES.md

@@ -1795,7 +1795,8 @@ until a string is parsed.
 | | 1.3.0 | 1.4.0 onwards |
 |---|---|---|
 | `integral(f, x, 1)` | accepted, read as `integral(f, x)` | `FunctionArgumentCountException` |
-| `integral(f, x, a, b)` | — | the definite integral from `a` to `b` |
+| `integral(x, x, 2)` | the second antiderivative, `x ^ 3 / 6` | `FunctionArgumentCountException` |
+| `integral(x, x, 0, 1)` | `FunctionArgumentCountException` | the definite integral from `0` to `1` |
 
 The third argument used to be a repetition count and is now the lower bound of a range, so three
 arguments name neither form and are refused. `derivative(f, x, n)` is unaffected and still takes an
@@ -1808,7 +1809,11 @@ against 1.4.0 since January without anyone seeing it, because they were pinned t
 [#846](https://github.com/asc-community/AngouriMath/pull/846) is what surfaced it.
 
 **What to do.** `integral(f, x, 1)` becomes `integral(f, x)` — identical in meaning, since 1.3.0
-turned the former into the latter anyway. For a repeated antiderivative, nest the calls.
+turned the former into the latter anyway. A count of two or more was not a no-op and is the part
+that is actually gone: nest the calls instead, `integral(integral(f, x), x)`, which 1.3.0 accepted
+as well. Expect the answer to differ by the constants of integration, which the same PR began
+adding — where `integral(x, x, 2)` gave `x ^ 3 / 6` on 1.3.0, the nested form gives
+`C_1 + C * x + x ^ 3 / 6` from 1.4.0 onwards.
 
 ---