Define hash for the remaining spaces

[jishnub]

Follow-up to #698, which defined hash for PointSpace and DiracSpace. Every other space in this package was in the same position: Space.jl:165 defines == as spacescompatible plus equal domains, while hash fell back to the identity-based default, so equal-valued spaces hashed differently whenever they held a mutable field, or whenever their domains compared equal without being identical objects.

julia> u1, u2 = UnionDomain([0..1, 2..3]), UnionDomain([2..3, 0..1]);

julia> ConstantSpace(u1) == ConstantSpace(u2)
true

julia> hash(ConstantSpace(u1)) == hash(ConstantSpace(u2))
false

Since == is a conjunction that includes domain(A) == domain(B), hashing the domain always agrees with ==, and this is all that is needed for the spaces that only compare domains: ZeroSpace, ConstantSpace, ContinuousSpace and SplineSpace (whose order is mixed in, so that splines of different orders over the same domain don't collide). The others additionally hash whatever spacescompatible compares:

space	hashed
ZeroSpace, ConstantSpace, ContinuousSpace, SplineSpace{λ}	domain (and λ)
SumSpace, PiecewiseSpace, ArraySpace, TensorSpace	components
SubSpace	space and indices
QuotientSpace	the underlying space, since the boundary conditions are not compared by == either
PiecewiseSegment	points

Two direct sums may hold their pieces in different containers and still be equal (#574), so their components are hashed one at a time rather than hashing the container. Base hashes a range through its length, which it can't do for the infinite ranges that SubSpace often uses, so those are hashed through their first index and step instead.

PiecewiseSegment is a domain rather than a space, but it overrides == to compare points and has the same missing hash; ContinuousSpace and SplineSpace hash it as their domain.

Two things deliberately left out:

• Spaces that compare their domains do so through domainscompatible, which uses isapprox. == is therefore not transitive for them and no hash can respect it exactly. Hashing the domain reproduces == for exactly equal domains, which is the case that matters for a Dict or a Set.
• Fun and Segment have the same mismatch (Fun compares coefficients and space, Segment compares endpoints across types), and ProductSpace/PathologicalQuotientSpace define no spacescompatible at all, so == errors for them. These seem worth separate PRs.

The version bump is a separate commit, in case you'd rather batch it.

Tests: the full suite passes locally, with a new hash testset in test/SpacesTest.jl covering each type above.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Ay3nR9CQtrrZj3r5LFu59c

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 74.39%. Comparing base (023b1a4) to head (d8a8fdd).
⚠️ Report is 50 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #700      +/-   ##
==========================================
- Coverage   75.20%   74.39%   -0.81%     
==========================================
  Files          79       81       +2     
  Lines        8445     8659     +214     
==========================================
+ Hits         6351     6442      +91     
- Misses       2094     2217     +123     

☔ View full report in Codecov by Harness.
📢 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.