Support stieltjes/cauchy of ChebyshevT via Legendre

[dlfivefifty]

This drops the explicit-formula approach. stieltjes(J::AbstractJacobi, z) = stieltjes(Legendre(), z) * (Legendre() \ J) already covers ChebyshevT, and cauchy divides that lazy product by -2πi. The problem was that indexing rows of the product, e.g. stieltjes(ChebyshevT(), 2+im)[1,1:10] or anything from cauchy(ChebyshevT(), z), stalled inside LazyArrays. The fixes are upstream:

• transpose/adjoint of inv(UpperTriangular(...)) keep the triangular structure (JuliaArrays/LazyArrays.jl@913b6ac). The row indexing of a product transposes its factors, so this gives it the correct rowsupport.
• A block of a triangular inverse is computed from the principal block (also 913b6ac).
• A 1×∞ DualLayout lazy product broadcasts lazily, so cauchy's scalar division no longer materialises (913b6ac).
• Scalar indexing of an inverse lower-triangular matrix, and multiplication by blocks of one, are finite (Finite indexing of inverse lower triangular matrices and their blocks JuliaArrays/LazyArrays.jl#429).

With those in place, cauchy(ChebyshevT(), 2+im)[1,1:1000] takes about 0.05s.

Changes in this PR:

• Tests for stieltjes/cauchy of ChebyshevT: against quadrature, against the inhomogeneous three-term recurrence, on the interval with ±0im, for expansions, for vectors of points, and for mapped chebyshevt(0..2).
• The StieltjesPoints * Legendre simplify rule becomes StieltjesPoints * AbstractJacobi, so inv.(zs .- x') * ChebyshevT() works too.
• LazyArrays = "2.14.3", and the version is bumped to 0.4.2.

CI will fail until LazyArrays 2.14.3 is released with both fixes. The full test suite passes locally with those fixes applied to LazyArrays v2.14.1.

Still not working through this route: ChebyshevU/Ultraspherical, because ClassicalOrthogonalPolynomials has no Legendre \ ChebyshevU conversion, and general Jacobi, because Legendre \ Jacobi(a,b) is the inverse of a banded product and its triangular structure isn't recognised.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 95.83333% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 92.32%. Comparing base (c0c2796) to head (69c7331).

Files with missing lines	Patch %	Lines
src/stieltjes.jl	95.83%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #68      +/-   ##
==========================================
+ Coverage   91.79%   92.32%   +0.53%     
==========================================
  Files           4        4              
  Lines         317      365      +48     
==========================================
+ Hits          291      337      +46     
- Misses         26       28       +2     

☔ 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.