Large memory allocation in stieltjes, backwardrecurrence!

[jamiecjx]

stieltjes() evaluated at certain points results in large amount of memory allocated.

julia> using ClassicalOrthogonalPolynomials, SingularIntegrals

julia> W = Weighted(ChebyshevU()); x = axes(W, 1); f = expand(W, x -> inv(√2 * x + 3/2) * √(1-x^2));

julia> z = 0.2567881003580743 - 0.33437737333561895im;

julia> @time inv.(z .- x') * f # first time run, ignore
  1.994925 seconds (1.61 M allocations: 3.126 GiB, 10.78% gc time, 65.63% compilation time)
1.1162556659904905 + 1.449666329259747im

julia> @time inv.(z .- x') * f # large memory allocation
  0.614899 seconds (78 allocations: 2.980 GiB, 19.81% gc time)
1.1162556659904905 + 1.449666329259747im

julia> @time inv.(z+0.01im .- x') * f # fine here
  0.000219 seconds (80 allocations: 122.562 KiB)
1.1295620627968406 + 1.4556063512794264im

After some investigation, it seems that the problem resides in backwardrecurrence!, where:

Line 118: μ is set to 1
Line 121, N is set to maxiterations, which results in K.data being 100_000_000 size.

Setting maxiterations to something smaller seems to "fix" the problem.

[dlfivefifty]

Do you think you could make a PR? Then I can code review it