Fix planar index reordering and remove _contractedspace

[lkdvos]

Fixes the root cause that #531 works around, and supersedes it.

Problem

@planar allocates through TensorOperations.tensoralloc_contract with the raw index tuples of the planar decomposition, in which pA and pB need not be planar partitions by themselves — only planarcontract! rotated them into shape, and only when pAB could be absorbed into those rotations. Since #515 added the coloring checks to the ProductSpace/HomSpace constructors, computing the output space as permute(compose(permute(VA, pA), permute(VB, pB)), pAB) throws for GenericUnit sectors, which is why #515 replaced it by the leg-wise _contractedspace. That workaround pins both operands to a single spacetype, which breaks contracting operands whose spacetype parameters differ (a BlockTensorMap with a TensorMap, as reported from MPSKit in #531); the fallback added there fixes the MethodError but reintroduces exactly the intermediate spaces that #515 removed.

Changes

• planarcontract! takes an arbitrary cyclic pAB, like the non-planar contract!: a residual permutation is applied with transpose! through an intermediate, mirroring what blas_contract! does with permute!. The remaining difference between planar and non-planar contraction is transpose vs permute, which is also what the AD rules of Planar additions [WIP] #124 need.
• reorder_indices is replaced by planar_contract_indices(A, pA, B, pB, pAB) -> pA′, pB′, pAB′: it rotates each operand's index cycle so the contracted indices form a contiguous arc and remaps pAB instead of absorbing it. It is shared with the two BraidingTensor methods and works on HomSpaces as well as on tensors.
• @planar canonicalizes the index tuples it emits at macro-expansion time: planar_contract_indices only needs the index partitions of both operands, and those are written in the expression and already checked against the actual tensors by _extract_tensormap_objects. Both the contraction and its allocation therefore see planar partitions, and the allocation needs no planar-specific entry point. Hand-written planar kernels should canonicalize with planar_contract_indices before allocating with tensoralloc_contract; planarcontract! keeps canonicalizing at runtime, so the kernel itself is safe either way.
• ⊗ allocates its destination directly; its pA = ((all indices), ()) intermediate is non-planar by construction.
• tensorcontract_structure composes the intermediate spaces again and _contractedspace is gone, which restores mixed spacetype support. For GenericUnit sectors, pA and pB passed to it must now be planar partitions.

Tests

New multifusion coverage (there was no @planar binary contraction on multifusion spaces): a planar contraction whose partitions are non-planar by themselves, ⊗ with more than one index in the domain, a non-trivial and a non-cyclic pAB, planar_contract_indices unit tests including the fully contracted and non-planar cases, and a check on the index tuples of the macro expansion.

Verified on the full space list for test/tensors/{contractions,braidingtensor,planar,linalg,indexmanipulations}.jl and test/symmetries/spaces.jl. The mixed-spacetype case has no in-repo coverage (it needs a downstream space type), so it is worth a run against BlockTensorKit + MPSKit before merging.

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 97.75281% with 2 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/planar/planaroperations.jl	95.55%	2 Missing ⚠️

Files with missing lines	Coverage Δ
src/planar/macros.jl	84.94% <100.00%> (+0.50%)	⬆️
src/planar/postprocessors.jl	100.00% <100.00%> (ø)
src/planar/preprocessors.jl	88.43% <100.00%> (+0.21%)	⬆️
src/spaces/homspace.jl	92.25% <100.00%> (-0.43%)	⬇️
src/tensors/braidingtensor.jl	88.05% <100.00%> (+0.46%)	⬆️
src/tensors/linalg.jl	83.23% <100.00%> (+0.30%)	⬆️
src/tensors/tensoroperations.jl	96.44% <100.00%> (ø)
src/planar/planaroperations.jl	70.31% <95.55%> (-2.88%)	⬇️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

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

[leburgel]

Always happy to be superseded, this looks great!

[Jutho]

Up to small suggestions, very much approved.