fix(engine): parse Land Equilibrium's chained control-gated sacrifice replacement

[rykerwilliams]

Summary

Land Equilibrium ({2}{U}{U}, Enchantment, Legends): "If an opponent who controls at least as many lands as you do would put a land onto the battlefield, that player instead puts that land onto the battlefield then sacrifices a land of their choice."

Previously parsed with zero coverage — the chained "then sacrifices" clause was swallowed (SwallowedClause/Replacement_Instead parse warning, replacements: []).

• Adds a nom combinator for the "an opponent who controls at least as many <X> as you do" subject-relative clause (CR 614.1a), reusing the existing comparator-agnostic player_count_comparison_filters helper rather than the existential-aggregate shape used by "an opponent controls more lands than you."
• Wires this into a new ReplacementDefinition that gates on the entering opponent's own land count — not an aggregate check across all opponents — via a new ScopedPlayer-aware QuantityContext threaded through OnlyIfQuantity/UnlessQuantity condition evaluation (game/replacement.rs). Verified this is additive: none of the 27 existing cards using these condition variants reference ScopedPlayer today.
• Generalizes the sacrifice-execute and event-tag derivation from the same parsed type filter rather than hardcoding "land" in three places, so the whole replacement (condition + event-tag + execute) stays internally consistent if a sibling card with a different permanent noun is ever found; a mismatch fails safely (None, falls through to the next dispatch branch) rather than misparsing.
• Fixes a latent continuation-source bug in stash_post_replacement_continuation: the stashed source used rid.source (the replacement's bearer) instead of the entering object. This is a no-op for every existing card reaching this code path (all 21 Devour-family cards have valid_card: SelfRef, so bearer == entering object by construction — verified via a whole-database query), but is required for Land Equilibrium's TokenEntry drain path (an opponent creating a land token under Land Equilibrium), since it's the first external (non-self-referential) replacement to reach this code.

Test plan

• New parser unit test: Land Equilibrium's real Oracle text produces exactly one Moved replacement with OnlyIfQuantity{GE} and a Sacrifice execute (no SwallowedClause), asserting exact target-filter equality.
• New integration test suite (land_equilibrium_forced_sacrifice.rs, 5 scenarios), driving the real GameRunner/PlayLand pipeline:
  • Opponent with more lands than caster is forced to sacrifice (land still enters first).
  • Opponent with fewer lands plays normally, no sacrifice.
  • Equal counts still trigger (GE boundary).
  • 3-player fixture proves the gate binds to the specific entering opponent, not an aggregate check (one gated opponent, one not, despite the caster's own count being fixed).
  • An existing Devour-family card's post-replacement continuation is unaffected by the stash-source change (regression guard).
• cargo clippy -p engine --tests -- -D warnings clean
• cargo fmt --all
• Parser combinator gate (./scripts/check-parser-combinators.sh) clean
• All CR citations (614.1a, 614.1c, 614.13, 608.2c, 109.5, 701.21a) verified against the Comprehensive Rules text

🤖 Generated with Claude Code

https://claude.ai/code/session_01XbgwGxbU9NHN9kou9isp8K

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[github-actions]

Parse changes introduced by this PR · 1 card(s), 2 signature(s) (baseline: main a61e7fdd919c)

1 card(s) · ability/ChangeZone · removed: ChangeZone (target=parent target, to=battlefield)

Examples: Land Equilibrium

1 card(s) · replacement/Moved · added: Moved

Examples: Land Equilibrium

2 card(s) had Oracle-text changes (errata/reprint) — excluded as non-parser.

[matthewevans]

Approved — right seam, verified rules evidence, and a surgical one-card parse-diff. Held last sweep only for missing evidence; both gaps have now closed favorably.

✅ Clean

Right seam — the combinator rejects the shape that would have been subtly wrong. The new nom combinator reuses player_count_comparison_filters and explicitly declines the existential-aggregate PlayerFilter::ControlsCount shape. That matters: an aggregate filter would answer "does some player control more lands?" when the rule needs "does this entering opponent control more lands?" — it would pass casual tests and bind the wrong player in a multiplayer game. Rejecting it is the harder and correct choice.

Shared-seam change is correctly scoped. rid.source → new_event.affected_object_id().unwrap_or(rid.source) touches the Applied arm only. Leaving the Prevented arm alone is right — a prevented event has no entering object to bind to — and the unwrap_or fallback plus the Devour-family regression test guard the blast radius.

Evidence, verified rather than taken on trust:

• Oracle text matches Scryfall exactly (checked against the card, not the PR description).
• All 6 CR citations grep-verify against docs/MagicCompRules.txt: 614.1a, 614.1c, 614.13, 608.2c, 109.5, 701.21a.
• Parse-diff: exactly 1 card, 2 signatures. Land Equilibrium loses a bogus ChangeZone (target=parent target, to=battlefield) and gains replacement/Moved. Zero collateral anywhere else in the pool — precisely the claimed scope, which is the signature you want from a replacement-effect fix.
• Tests are registered in tests/integration/main.rs (not inert) and discriminating: three_player_gate_binds_to_specific_entering_player would fail against an aggregate implementation.

CI is genuinely green — both Rust shards ran real work (9m14s / 9m15s), lint 8m18s, card-data 10m21s. Not a hollow pass.

Recommendation: approve and enqueue. Nothing blocking.