introduce fast mode for optimized sibling presence checks

Centralize the layer count validation into PartitionCache and track it via a new fast_mode flag. Extend query tuples to include a pre-resolved destination layer index, enabling a fast-path batch lookup that bypasses per-layer probing when enabled. Refactor neighbor iteration and hit resolution to eliminate duplication and conditionally dispatch to the optimized path based on the cache state.

src/obikphylo/src/siblings/build.rs

@@ -132,20 +132,12 @@ fn build_layer_sibling_annex(
     let n = mphf.n();
 
     // Whether this layer's family-member fields can carry a real layer
-    // number (fast path, exploited by a future scan-time reader) or must
-    // stay presence-only (today's behaviour) — a single fact about the
-    // whole index, decided once here, never stored: `n_layers` is
-    // guaranteed identical across every partition (a merge adds one layer
-    // to all of them at once), and `FamilyMask`'s 3-bit field only has room
-    // for layers `0..=6`.
-    let fast_mode = meta.n_layers <= 7;
-    if !fast_mode {
-        tracing::warn!(
-            "layer {l} ({layer_dir:?}): index has {} layers (>7) — sibling-annex fast layer \
-             lookup disabled for this build; consider compacting this index",
-            meta.n_layers
-        );
-    }
+    // number (exploited by `family_scan::scan_layer_families`'s fast path)
+    // or must stay presence-only — a single fact about the whole index,
+    // decided once by `cache` (built from the same `PartitionMeta` this
+    // function would otherwise re-derive) so the writer and every reader
+    // agree; see `PartitionCache::fast_mode`'s docs.
+    let fast_mode = cache.fast_mode();
 
         // ── The annex file itself is the reconciliation state, indexed by
         // this layer's k-mer iteration order (the physical layout of