benchmarks: Simplify reading Zlib input streams

Author: freya022

benchmarks/src/jmh/java/dev/freya02/discord/zstd/ZstdStreamingBenchmark.java

@@ -121,22 +121,17 @@ public void zlib(ZlibDecompressorState decompressorState, ShardsState shardsStat
     @Benchmark
     public void zlibNoDeser(ZlibDecompressorState decompressorState, ShardsState shardsState, Blackhole blackhole) throws IOException {
         var decompressor = decompressorState.decompressor;
-        var bytes = decompressorState.buf;
         for (List<TestChunks.Chunk> shard : shardsState.shards) {
             decompressor.reset();
             for (TestChunks.Chunk chunk : shard) {
-                int currentlyDecompressedSize = 0;
-                int expectedDecompressedSize = chunk.decompressed().length;
                 try (InputStream inputStream = decompressor.createInputStream(chunk.zlibCompressed())) {
-                    // This is pretty stupid, #available() returns 1 even when there is no output to be read,
-                    // we want to avoid handling EOFException as it may be slow and does not represent real world usage,
-                    // checking `read < buf.length` is not viable since it can store data internally and returned in the next call.
-                    // So, we instead decompress until we have the known decompressed data length.
-                    do {
-                        var read = inputStream.read(bytes);
-                        currentlyDecompressedSize += read;
-                        blackhole.consume(bytes);
-                    } while (currentlyDecompressedSize < expectedDecompressedSize);
+                    while (true) {
+                        var read = inputStream.read(decompressorState.buf);
+                        blackhole.consume(decompressorState.buf);
+                        if (read <= 0) {
+                            break;
+                        }
+                    }
                 }
             }
         }