test: tighten the five Risch difficult-baseline entries the engine now solves

[s-celles-bot]

test/difficult_baseline.jl says "Engine improvements should tighten the codes here on the
same PR." Five Risch entries have been reporting as improvements for a while without being
tightened, so every difficult job prints:

Risch improvements vs baseline (engine got better — tighten test/difficult_baseline.jl on this PR):
  [3] sin(x)              expected=1 actual=0
  [26] sin(x)*(x^2)       expected=1 actual=0
  [27] cos(x)*(x^3)       expected=1 actual=0
  [49] sin(5x)*(2 + 3x)   expected=1 actual=0
  [55] sin(1 / x) / (x^2) expected=1 actual=0

Risch solves all five and verifies symbolically against the reference, i.e. code 0, where the
baseline still expects code 1. This updates the second element of those five tuples. The gate
is isempty(regressions), so nothing was failing — but the entries cannot catch a future
regression while they are loose, which is the point of the file.

Evidence that these are platform-independent

The five appear in every difficult job, so tightening is safe across the matrix:

• all 9 jobs (3 OS x Julia 1.10 / 1 / pre) of the most recent run on main and on a
  current pull request branch;
• locally on Julia 1.13.0 under SymbolicUtils 4.49.0 and 4.50.0, identically.

What this deliberately does not touch

The RuleBased improvements reported alongside them are not platform-independent, so they
stay loose:

entry	jobs reporting an improvement
[6] (x^2)*sqrt(1 + x)	4 of 9 — Julia 1.10 ubuntu, and macOS on all three versions
[129] 1 / (-x + x^3)	same 4 of 9
[133] (-3 + x) / (2x + 3(x^2) + x^3)	same 4 of 9
[132] 1 / ((4 - 4x + x^2)*(5 - 4x + x^2))	5 of 9 — Windows on all three versions, plus Julia 1.10 ubuntu and macOS

Note that the [6]/[129]/[133] set and the [132] set are almost disjoint by OS, which is
what entry 6's existing comment already records ("RB matches reference exactly on Julia 1.10
ubuntu/macOS but produces an equivalent unverifiable form (code 1) on Julia 1.x and
Windows"). Tightening any of them would turn a green matrix red on the other platforms.

Verification

TEST_GROUP=difficult with the tightened baseline: no regressions, and the five lines are
gone from the improvements report. Julia 1.13.0, SymbolicUtils 4.50.0.

Assisted by: AI

[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 51.19%. Comparing base (8647f0d) to head (5306aef).
⚠️ Report is 27 commits behind head on main.
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #153      +/-   ##
==========================================
+ Coverage   50.98%   51.19%   +0.21%     
==========================================
  Files          23       23              
  Lines        4309     4213      -96     
==========================================
- Hits         2197     2157      -40     
+ Misses       2112     2056      -56     

☔ 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.