Stop CreateUnique from handing back a name already in use, fixing wrong integrals

[Rafael-SOWNet]

A wrong answer in integration, and its cause is a variable-naming collision rather than
anything mathematical.

∫ x·(x²+1)² dx     // at x = 1.3 the integrand is 9.4069
                   // the derivative of the answer is 10.1649

Not in the tracker. I found it by sweeping integrands numerically — differentiating each
answer back and comparing at points — after a related wrong answer turned up in a case
the test suite was asserting.

What happens

Substitution asks Variable.CreateUnique for a variable that does not occur in the
expression. CreateUnique worked out which names to avoid via SplitIndex, which cuts
at the first underscore. For the prefix "u_sub" — the one integration uses — it read
an existing u_sub_1 as prefix "u" and index "sub_1", parsed no number out of it,
recorded nothing to avoid, and returned u_sub_1.

So the "fresh" variable was the variable being integrated. This only bites when
integration substitutes twice, because the second call is the first whose integration
variable is itself a u_sub.

The consequence is quiet. Substituting u_sub_1^2 → u_sub_1 turns

(u²/2 + u) / (2u)      into      (3/2·u) / (2u)  =  3/4

A constant — containing no trace of the variable being integrated, which is exactly the
test for "the substitution worked". A nonsense expression was accepted and integrated,
and the answer looked plausible.

The change

CreateUnique compares against the names actually in use rather than against indices
parsed out of them, so it no longer depends on how SplitIndex splits anything. Any
prefix containing an underscore was affected; "u_sub" is the one that reaches users.

Result

Three existing tests change because the answers do. Each new answer was verified by
differentiating back at several points:

	was	now
∫ x(x²+1)³ dx	the expanded polynomial	(x² + 1)⁴ / 8
∫ x(x²+1)¹⁰ dx	the expanded polynomial	(x² + 1)¹¹ / 22
∫ 3x²(x³+2)² dx	not an antiderivative	(x³ + 2)³ / 3

The second one is worth a look: its old expected value carried the comment
"differs from C + (x^2+1)^11 / 22 by -1/22". That is now simply the answer.

Two integrals that used to come back unevaluated — x(x²+x+1)² and x²(x²+1)² — are
solved as well.

Testing

UnitTests 3672 passed / 0 failed, FSharpWrapperUnitTests 127 passed / 0 failed, and
both netstandard2.0 and net7.0 build.

14 new tests: the two wrong answers, the two that were unsolved, the closed forms, and
the plain single-level substitutions that already worked and must keep working. They
assert by differentiating back rather than against printed strings, since pinning the
string is what let the original wrong answer sit in the suite.

Independent of #663 through #669.

🤖 Generated with Claude Code

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 79.44%. Comparing base (90c00a8) to head (8a2d90c).
⚠️ Report is 41 commits behind head on master.
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #670      +/-   ##
==========================================
- Coverage   80.99%   79.44%   -1.56%     
==========================================
  Files         155      155              
  Lines       13687    12291    -1396     
  Branches     1957     1980      +23     
==========================================
- Hits        11086     9764    -1322     
+ Misses       1990     1941      -49     
+ Partials      611      586      -25     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.