docs: link explanation.md to the ADRs

Cross-reference ADR 0001 (FFM over JNI) and ADR 0002 (zig cc) inline in the
Why FFM and Zig section, and add an Architecture decisions section pointing at
the ADR index.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: dfa1

docs/explanation.md

@@ -5,13 +5,15 @@
 The bindings use the **Foreign Function & Memory API** rather than JNI: no
 hand-written C glue, no separate native compile step for the binding layer, and a
 direct path from Java to zstd's addresses — which is what makes the zero-copy
-`MemorySegment` API possible.
+`MemorySegment` API possible. See
+[ADR 0001 — FFM over JNI](../adr/0001-ffm-over-jni.md).
 
 The native library itself is built from vendored zstd source via **`zig cc`** as
 a drop-in C compiler. zstd is pure C with no build-system dependencies, so the
 sources are compiled directly — no autotools, no CMake. Zig bundles clang and
 libc for every target, enabling hermetic cross-compilation without a sysroot:
-any host can build all six platform artifacts.
+any host can build all six platform artifacts. See
+[ADR 0002 — zig cc native build](../adr/0002-zig-cc-native-build.md).
 
 ## When zero-copy pays off
 
@@ -23,3 +25,9 @@ does not matter, is in [zero-copy.md](zero-copy.md).
 
 Throughput and allocation versus zstd-jni (JNI) and aircompressor (pure Java),
 including an async-profiler breakdown: [benchmarks.md](benchmarks.md).
+
+## Architecture decisions
+
+The reasoning above is distilled from the full set of Architecture Decision
+Records — every foundational choice, its alternatives, and its trade-offs, one
+file per decision: [adr/ADR.md](../adr/ADR.md).