perf: reduce contraction materialization overhead

[GiggleLiu]

Summary

• add layout-aware Standard GEMM fast paths for unbatched and batched CPU contractions
• reuse scratch buffers during contraction materialization and strengthen CPU/backend regression coverage
• let optimized roots emit final output order directly when finalization would only permute indices

Test Plan

• make check
• targeted regression tests from the implementation plan
• cargo bench --bench binary (repo has no bench target named binary)
• make bench-network (repo has no bench-network target)
• python3 benchmarks/compare.py (benchmarks/compare.py is absent)
• samply record cargo run --release --example profile_network -- --scenario 3reg_150 --iterations 1 --output /tmp/omeinsum-profile-network-after.json (repo has no profile_network example)

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 93.13929% with 33 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
src/backend/cpu/contract.rs	92.56%	20 Missing ⚠️
src/einsum/engine.rs	77.08%	11 Missing ⚠️
src/backend/cpu/mod.rs	98.33%	2 Missing ⚠️

📢 Thoughts on this report? Let us know!