Add growth state frontmatter to all 280 wiki files

TREE(74): training modules, entity profiles, 2890 references, keyword indices
SPROUT(42): knowledge pages, project docs, curated source material
SEED(164): daily notes, raw session logs, unprocessed material

Updated AUDIT_MANIFEST.json with growth classifications.

memory/2026-05-04.md

@@ -1,3 +1,6 @@
+---
+growth: seed
+---
 # 2026-05-04 — Memory Flush (Pre-Compaction)
 
 ## Session Start — Swerve Hardware Specs (00:52 UTC)

memory/connections-log.md

@@ -1,3 +1,6 @@
+---
+growth: seed
+---
 ## 2026-05-04 14:31 UTC
 Connections found: 1
 - [HHS-Hackers] "[[M5 + Thunderbolt scaling]]" ↔ [Research] "[[MLX Distributed Inference]]" — Chris asked crash-bot about GPU scaling options, was directed to MLX. Key finding: MLX supports true distributed inference where model layers split across multiple Apple Silicon machines, both working on every token simultaneously. This is the "uber geek play" for scaling — not just two separate machines but a unified inference cluster via Thunderbolt. Relevant to Chris's long-term AI rig planning.

memory/research-queue.md

@@ -1,3 +1,6 @@
+---
+growth: seed
+---
 # Research Queue — Fabric Growth
 
 ## Active Queue