Factored RMDPs as a base model

[Zinoex]

... and customizable Bellman algorithms, inferred model types (influences default Bellman algorithm), update documentation including add DocumenterCitations, pretty-print models, parameterize tests.

Fix #84

[Copilot]

Pull Request Overview

Adds support for factored / robust (ambiguity-set based) interval MDP models by replacing IntervalProbabilities with IntervalAmbiguitySets across the test suite, introduces extensive new tests for factored, threaded, vertex-enumeration, McCormick, and O-max Bellman/backward operators (including GPU variants), and updates specification / product / IO tests to reflect the new abstractions. Also adds convenience solver return-object assertions and implicit sink‑state scenarios while removing the previous orthogonal probability test file.

• Migrate tests from IntervalProbabilities to IntervalAmbiguitySets (including API adjustments).
• Add comprehensive factored / ambiguity set Bellman and value iteration test coverage (CPU & CUDA).
• Introduce solver result accessors and implicit sink state tests; remove outdated orthogonal tests.

Reviewed Changes

Copilot reviewed 85 out of 89 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
test/sparse/synthesis.jl	Updated synthesis tests to new ambiguity set API; added solution object assertions and implicit sink state; potential variable shadowing issue.
test/sparse/sparse.jl	Replaced orthogonal test reference with new factored test file.
test/sparse/orthogonal.jl	Removed (deprecated orthogonal probability tests).
test/sparse/imdp.jl	Migrated to ambiguity sets; broadened numeric type coverage; added many additional property tests.
test/sparse/factored.jl	New extensive factored robust MDP Bellman / strategy tests (McCormick, OMax, vertex, threaded).
test/sparse/bellman.jl	Refactored Bellman tests to ambiguity sets & new workspace types.
test/data/prism.jl	Adjusted model IO tests to ambiguity set / marginal accessors.
test/data/intervalmdp.jl	Updated model IO round‑trip tests to ambiguity sets.
test/data/bmdp_tool.jl	Updated BMDP tool IO tests; added test grouping and ambiguity set assertions.
test/cuda/sparse/vi.jl	Ported sparse GPU VI tests to ambiguity sets with richer assertions.
test/cuda/sparse/synthesis.jl	CUDA synthesis tests updated; added solution object assertions.
test/cuda/sparse/imdp.jl	GPU IMDP tests expanded (exact/finite/infinite properties) using ambiguity sets.
test/cuda/sparse/bellman.jl	Adjusted GPU sparse Bellman tests; increased stress test sizes (possible perf concern).
test/cuda/dense/vi.jl	Dense GPU VI tests migrated to ambiguity sets & extended.
test/cuda/dense/synthesis.jl	Dense GPU synthesis tests updated with solution object checks.
test/cuda/dense/imdp.jl	Dense GPU IMDP tests expanded similarly to sparse versions.
test/cuda/dense/bellman.jl	Dense GPU Bellman tests refactored to new API.
test/base/vi.jl	Base (CPU) VI tests updated; broadened numeric types; added bounds checks.
test/base/synthesis.jl	Base synthesis tests updated; added solution accessor assertions; variable shadowing issue.
test/base/specification.jl	Specification getter/error tests updated (qualified isfinitetime); adjusted strategies.
test/base/product.jl	Product model tests adapted to ambiguity sets; removed 3rd (sink-style) probability in some cases.

Comments suppressed due to low confidence (1)

test/base/synthesis.jl:63

• Same variable shadowing issue: the loop rebinds k so num_iterations(sol) is compared to the loop's final index instead of the iteration count returned by solve. Rename the loop variable (e.g. t) or preserve the original k in another identifier before the loop.

@test policy isa TimeVaryingStrategy
@test time_length(policy) == 10
for k in 1:time_length(policy)
    @test policy[k] == [(1,), (2,), (1,)]
end

@test strategy(sol) == policy
@test value_function(sol) == V
@test num_iterations(sol) == k

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}