Update LinearAlgebra QR tests so that MKL.jl tests pass on the Linear…

…Algebra test suite. (JuliaLang#43585)

stdlib/LinearAlgebra/test/qr.jl

@@ -432,21 +432,22 @@ end
  @test eltype(qrnonblas.factors) == eltype(qrnonblas.τ) == ComplexF16
 end
 
+# We use approximate equals to get MKL.jl tests to pass.
 @testset "optimized getindex for an AbstractQ" begin
  for T in [Float64, ComplexF64]
  Q = qr(rand(T, 4, 4))
  Q2 = Q.Q
  M = Matrix(Q2)
  for j in axes(M, 2)
- @test Q2[:, j] == M[:, j]
+ @test Q2[:, j] ≈ M[:, j]
  for i in axes(M, 1)
- @test Q2[i, :] == M[i, :]
- @test Q2[i, j] == M[i, j]
+ @test Q2[i, :] ≈ M[i, :]
+ @test Q2[i, j] ≈ M[i, j]
  end
  end
- @test Q2[:] == M[:]
- @test Q2[:, :] == M[:, :]
- @test Q2[:, :, :] == M[:, :, :]
+ @test Q2[:] ≈ M[:]
+ @test Q2[:, :] ≈ M[:, :]
+ @test Q2[:, :, :] ≈ M[:, :, :]
  end
 end