Complete GC improvements: implement pointer tagging, vector pooling, dedup scratch, and recursive clone optimization

Implement all 8 identified GC optimizations: pointer tagging infrastructure (SmallInt/Bool/Unit encoding with rotating unpack scratch), pinned-chunk pooling with madvise reclamation, record vector pooling (7-bin cache), thread-local dedup scratch reuse, two-tier string interning (L1 thread-local + L2 global), iterative deep_clone_value (eliminates stack overflow on fork environments), and conditional GC stats counters (gated by gc-stats feature). All 422 tests pass in both release and gc-stats modes; 6 example programs verified working with tagging active.

Author: ilyakooo0

Cargo.lock

@@ -6,6 +6,13 @@ edition.workspace = true
 [lib]
 crate-type = ["staticlib", "rlib"]
 
+[features]
+default = []
+# Enable GC telemetry counters (fetch_add per allocation).  Off by default
+# in release for zero overhead; turn on for profiling via
+# `cargo build --release --features gc-stats`.
+gc-stats = []
+
 [dependencies]
 rusqlite = { workspace = true }
 tiny_http = { workspace = true }
@@ -20,3 +27,6 @@ x25519-dalek = { version = "2", features = ["static_secrets"] }
 chacha20poly1305 = "0.10"
 serde_json = { workspace = true }
 ureq = { workspace = true }
+
+[target.'cfg(unix)'.dependencies]
+libc = "0.2"