M11: Prove reservation-core as a third deterministic engine inside allocdb

[skel84]

Summary

Implement reservation-core as the third in-repo deterministic engine to pressure-test the library thesis after quota-core.

Why

quota-core proved that a second engine can live in the workspace, but the seam evaluation still concluded that shared runtime extraction is premature. reservation-core should apply stronger pressure to expiry, lifecycle, and terminal-state recovery seams.

Scope

• keep the experiment inside the current allocdb workspace
• add a sibling crates/reservation-core
• defer shared runtime extraction until after the third-engine readout

Canonical docs

• docs/reservation-engine-plan.md
• docs/reservation-semantics.md

Child tasks

• M11-T01 freeze reservation boundary and v1 scope
• M11-T02 scaffold reservation-core sibling crate with copied substrate
• M11-T03 implement CreatePool, PlaceHold, ConfirmHold, and ReleaseHold
• M11-T04 add logical-slot expiry and durability/recovery proof
• M11-T05 evaluate shared runtime seams after the third engine

[skel84]

M11 is complete via allocdb#115, allocdb#116, allocdb#117, allocdb#118, and allocdb#119.