fix: read through the facade in the test suite (closes #30)

[Dani-giron]

Summary

The apotheosis2 branch does not compile: the test suite added in #21 accesses Apotheosis::hnsw and Apotheosis::records, which #14 made private.

The root cause is that #14 was only partially implemented. Its Expected behavior specified draw_model(), len() and is_empty() on the facade; only the last two were added, so the tests had no facade method to read the graph structure through and reached into the private field instead.

Changes

• src/controllers/apotheosis.rs: adds the missing draw_model(), delegating to Hnsw::draw_model(). It adds no state and no new invariant: it exposes for reading the same data draw() already writes out as GEXF.
• The affected call sites move onto the facade: .hnsw.draw_model() becomes .draw_model() (13 sites) and .records.len() becomes the existing .len() (18 sites), across config_boundary.rs, config_matrix.rs, fitness_functions.rs, ann_correctness.rs and api_contract.rs. No test logic changed, only how each test reaches the data.

Test plan

Verified locally with the same commands CI runs: cargo clippy --all-targets --all-features -- -D warnings clean, cargo fmt --check clean, and all 44 tests passing.

Closes #30.

[Dani-giron]

Merge order note for the current batch of open PRs.

#34, #35, #37 were branched from this PR's branch (k1-facade-draw-model), so they already carry these commits forward. Merging this one first avoids any duplicate-commit confusion when those three land afterward, they'll simply recompute clean against the updated apotheosis2.

I looked into retargeting their base branch to this one to make GitHub enforce the order, but that only works when the base branch exists in this repo; k1-facade-draw-model only exists on my fork, so it's not possible for cross-fork PRs. Documenting it here instead.

Recommended order: this one (#31) first.