Optimize chunker hot loop for ~1.5-2x throughput

[folbricht]

Summary

• Replace expensive DIVL instruction in boundary detection (hValue % disc == disc-1) with Lemire's fast divisibility test — a multiply-and-compare that costs ~5 cycles vs ~26 for hardware division
• Convert hashTable from slice to [256]uint32 fixed-size array, eliminating bounds checks on all accesses
• Add precomputed hashTableRotated table, removing one RotateLeft32 call per byte in the hot loop
• Hoist struct fields (hValue, hIdx, buf, window pointer, Lemire constants) into local variables to avoid repeated pointer dereferences
• Replace (hIdx + 1) % 48 with a branch (if hIdx >= 48) — perfectly predicted, taken once per 48 iterations
• Reuse backing buffer in fillBuffer() to reduce allocations from O(filesize/buffersize) to O(1)
• Add b.SetBytes() to benchmarks for direct MB/s throughput reporting; modernize to use b.Loop()

All optimizations preserve identical chunk boundaries, verified by TestChunkerLargeFile which checks exact SHA512/256 hashes for every chunk.

Test plan

• go test -run TestChunkerLargeFile — exact chunk boundary regression (SHA512/256 hashes)
• go test -run TestChunker — all chunker tests pass
• go test -bench=BenchmarkChunk -benchmem — verify throughput improvement
• CI passes on Linux, macOS, Windows

🤖 Generated with Claude Code

Before:

goos: linux
goarch: amd64
pkg: github.com/folbricht/desync
cpu: Intel(R) Core(TM) i7-7600U CPU @ 2.80GHz
BenchmarkChunker-4         	     272	   4304447 ns/op	 2621489 B/op	       2 allocs/op
BenchmarkChunkNull1M-4     	     205	   5907772 ns/op	 2621489 B/op	       2 allocs/op
BenchmarkChunkNull10M-4    	      15	  69146171 ns/op	13107254 B/op	       6 allocs/op
BenchmarkChunkNull50M-4    	       4	 294544244 ns/op	55050312 B/op	      23 allocs/op
BenchmarkChunkNull100M-4   	       2	 593006637 ns/op	107479112 B/op	      43 allocs/op

After:

goos: linux
goarch: amd64
pkg: github.com/folbricht/desync
cpu: Intel(R) Core(TM) i7-7600U CPU @ 2.80GHz
BenchmarkChunker-4         	     472	   2371186 ns/op	 442.22 MB/s	 2621489 B/op	       2 allocs/op
BenchmarkChunkNull1M-4     	     376	   3125929 ns/op	 335.44 MB/s	 2621490 B/op	       2 allocs/op
BenchmarkChunkNull10M-4    	      38	  31382746 ns/op	 334.12 MB/s	 2621489 B/op	       2 allocs/op
BenchmarkChunkNull50M-4    	       7	 153479795 ns/op	 341.60 MB/s	 2621488 B/op	       2 allocs/op
BenchmarkChunkNull100M-4   	       4	 307347136 ns/op	 341.17 MB/s	 2621488 B/op	       2 allocs/op

Closes #244