feat(pathfinding): .swb format v7 per-chunk compression (#3a) (#2470)

## Summary
Phase #3a, stacked on #2469. Compresses each chunk record independently with libdeflate (random access preserved).

Trammel: 124.7 MB → 19.2 MB (−85%).

## Details
- Whole-record framing `[u32 UncompressedLen][payload]`; records that don't shrink (tiny Uniform) are stored raw, detected as payload length == UncompressedLen.
- Codec chosen by full-Trammel spike: libdeflate VeryHigh (16.5 MB, 1.83 µs/chunk decompress) over zstd L19/22 (17.4 MB) and managed Brotli q11 (16.4 MB) — best native ratio, fastest decompress, already the repo's packet codec (no new dependency).
- Reuses cached thread-static bindings: `Deflate.Maximum` for bake, `Deflate.Standard` for reads.
- Compression is bake-time only; decompression is one-time per chunk (LRU-cached).
- Format v7; v6 files rejected and re-baked once.

## Tests
v7 unit tests + the v6 suite run through the compression path; full pathfinding suite green; Release build clean.

Projects/Server/Compression/Deflate.cs

@@ -23,5 +23,13 @@ public static class Deflate
     [ThreadStatic]
     private static LibDeflateBinding _standard;
 
+    [ThreadStatic]
+    private static LibDeflateBinding _maximum;
+
     public static LibDeflateBinding Standard => _standard ??= new LibDeflateBinding();
+
+    // Best-ratio compressor. Construction allocates a native libdeflate compressor, so it is
+    // cached per thread like Standard and reused. Decompression is level-independent, so the
+    // decompress path can use either accessor. libdeflate is not thread-safe — hence ThreadStatic.
+    public static LibDeflateBinding Maximum => _maximum ??= new LibDeflateBinding(LibDeflateCompressionLevel.VeryHigh);
 }