Fix incorrect one-arg Base.hash for Block in NDTensors

[adienes]

Generated as part of an ecosystem-wide audit for one-arg hash methods.

Summary

hash(b::Block) in NDTensors/src/blocksparse/block.jl defines a one-arg Base.hash (via import Base: hash in imports.jl). This means the two-arg fallback hash(x, h::UInt) uses objectid-based hashing, which can cause:

• Correctness bugs: equal Blocks may hash differently when used as compound keys in Dict or Set
• Performance issues: excessive method invalidation across the ecosystem
• Julia 1.13+ breakage: the default hash seed is changing from zero(UInt) to a random value

The existing two-arg method also used addition (h + hash(b)) instead of proper hash chaining.

Fix

Replace both methods with a single proper two-arg hash(b::Block, h::UInt) = hash(b.hash, h) that chains the cached hash value through the seed.

Test plan

• No custom == for Block; default struct equality compares data and hash fields, which is consistent with the new hash method
• Block is used as Dict key (Dict{Block{NR}, Bool}), so correct Base.hash is important

---

This PR was generated with the assistance of generative AI.

Co-Authored-By: Claude noreply@anthropic.com

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 79.38%. Comparing base (294c359) to head (c33645e).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1716      +/-   ##
==========================================
- Coverage   81.09%   79.38%   -1.71%     
==========================================
  Files          59       53       -6     
  Lines        4676     4556     -120     
==========================================
- Hits         3792     3617     -175     
- Misses        884      939      +55     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

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

[mtfishman]

Thanks @adienes! Sorry for all the PR noise, we are reworking some of our CI workflows and this PR exposed an issue when running downstream tests of private repositories from fork PRs, looks like that is fixed now and tests are passing so I'll merge this.

[adienes]

glad it ended up helpful! no worries about the noise. I had ~30 of these PRs open 😂 so inbox was already cooked