Ensure no sharing of data in LQ gauge fixing

[lkdvos]

This fixes #218.

Somehow this slipped through the cracks, as I already addressed this issue for QR (

MatrixAlgebraKit.jl/src/implementations/qr.jl

Lines 178 to 182 in a53006e

	# we need to first copy then gaugefix - avoiding aliasing between R and Rd for broadcast
	Rd = diagview(A)
	Rf = pivoted ? view(R, :, jpvt) : R
	copyto!(Rf, uppertriangular!(view(A, axes(R)...)))
	positive && gaugefix!(qr_householder!, Q, Rf, Rd)

) but forgot to also add it in LQ.
The issue being that L .*= sign_safe.(diagview(L)) does not give correct results since L and diagview(L) share data in a non-trivial way.

Interestingly, broadcasting Base Array and views thereof detect this and make preventive copies, while JLArray does not have this behavior.
(I'm guessing there is some dataids overload missing or something for that to work):

julia> A = -jl(ones(3, 3))
3×3 JLArray{Float64, 2}:
 -1.0  -1.0  -1.0
 -1.0  -1.0  -1.0
 -1.0  -1.0  -1.0

julia> A .*= sign.(view(A, 1:4:9))
3×3 JLArray{Float64, 2}:
 1.0  -1.0  -1.0
 1.0   1.0  -1.0
 1.0   1.0   1.0

julia> B = -ones(3, 3)
3×3 Matrix{Float64}:
 -1.0  -1.0  -1.0
 -1.0  -1.0  -1.0
 -1.0  -1.0  -1.0

julia> B .*= sign.(view(B, 1:4:9))
3×3 Matrix{Float64}:
 1.0  1.0  1.0
 1.0  1.0  1.0
 1.0  1.0  1.0

In any case, this fix should now ensure correct results for JLArray's and one fewer allocation for regular Arrays.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

Files with missing lines	Coverage Δ
src/implementations/lq.jl	93.61% <100.00%> (ø)

🚀 New features to boost your workflow:

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