[codex] integrate upstream-safe stable surface

[MilkBlock]

Summary

This PR integrates the non-fork-dependent parts of the experimental branch onto stable while keeping the default branch green against upstream egglog.

The merged surface includes:

• relation and function query helpers
• display and typst metadata support
• action sample recorder examples and runtime hooks
• pseudo-singleton runtime/examples/tests
• benchmark CLI, progress, report, and example cleanup needed for a healthy default examples/test surface

What Was Blocking Stable

experimental had accumulated multiple features that assumed our custom egglog fork:

• proof-oriented rule metadata and pretty-proof helpers
• persisted snapshot / artifact / schema fingerprint paths
• timestamp-constrained rule attachment
• EBoost heuristic/layered extraction built on schema/raw introspection
• cached function-id fastpaths that assumed a root-level public egglog::FunctionId

Those assumptions prevented a direct move onto upstream egglog and made the default stable surface fail to compile or test cleanly.

What Changed

The stable integration branch now does three things:

1. It ports the upstream-safe feature groups from experimental.
   The query sugar, metadata support, recorder support, pseudo-singleton examples, and benchmark/example cleanup are all carried over.

2. It de-forks the runtime surface.
   The rule/runtime wrapper now uses upstream-compatible rust_rule paths by default, no longer requires fork-only rule metadata APIs, and no longer exposes fork-specific cached function IDs.

3. It explicitly fences off fork-only behavior.
   Proof, persisted-snapshot, artifact/schema, timestamp-filter attachment, and EBoost-specific paths are either feature-gated behind fork-egglog or downgraded to stable-safe fallback examples/stubs so the default branch does not pretend to support functionality that upstream egglog cannot provide.

User-Facing Effect

For stable reviewers and users, the default branch behavior is now coherent:

• cargo test passes on the stable integration branch using upstream egglog
• cargo build --examples passes
• the pseudo-singleton demos run successfully
• fork-only examples that cannot honestly work on upstream stable now fail soft with explicit guidance instead of compiling into broken codepaths

Validation

Fresh verification run on this branch:

• cargo test
• cargo build --examples
• cargo run --example constant_prop_pseudo_singleton
• cargo run --example constant_prop_pseudo_singleton_async
• cargo run --example relation_transfer_story
• cargo run --example timestamp_filter

Scope Boundaries

This PR intentionally does not make the following default-stable features available:

• proof-mode runtime support
• artifact / persisted snapshot / schema fingerprint support
• timestamp-constrained rule attachment on upstream stable
• EBoost heuristic/layered extraction on upstream stable

Those remain explicitly separated so we can review a truthful stable surface instead of a partial port that still depends on the custom fork.

[codspeed-hq]

Merging this PR will create unknown performance changes

🆕 1 new benchmark
⏩ 2 skipped benchmarks¹

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
🆕	eggplant_insert_1000_func	N/A	1.5 ms	N/A

---

_{Comparing stable-big-pr (20221d5) with stable (8902fb0)}

Footnotes

1. 2 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩