`unitary`/`isomorphism` produce a non-unitary map when the domain is a `SumSpace`

[VinceNeede]

Summary

Following up on this discussion: unitary/isomorphism produce a non-unitary map when the domain is a SumSpace, even though the total and per-sector dimensions match exactly between domain and codomain.

MWE

using TensorKit, BlockTensorKit, LinearAlgebra

V1 = Rep[U₁](0 => 2, 1 => 1)
V2 = Rep[U₁](0 => 2, 1 => 1)  
Vsum = V1 ⊞ V2
Vflat = oplus(Vsum)

@show dim(Vsum), dim(Vflat)                 # equal
for c in blocksectors(Vsum)
    @show c, dim(Vsum, c), dim(Vflat, c)    # equal, sector by sector
end

u = unitary(ComplexF64, Vflat, Vsum)
@show norm(u' * u - id(Vsum))               # expected ~0, got 1.7320508075688772

Expected behavior

norm(u' * u - id(Vsum)) should be ~0.

Actual behavior

Large nonzero residual — u has linearly dependent columns despite matching dimensions everywhere.