Fixup Brotli compression tests

[DrewScoggins]

As seen here, dotnet/runtime#73391 (comment), this test had a bug where in
the work that was being done by the underlying test was not being correctly measured. This fix resolves that.

[stephentoub]

It's less of an issue, but

performance/src/benchmarks/micro/libraries/System.IO.Compression/CompressionStreamPerfTestBase.cs

Line 71 in ccbeb4e

var compressor = CreateStream(CompressedFile.CompressedDataStream, CompressionMode.Decompress);

should also be disposing of this.

[stephentoub]

The same thing needs to be done in these:

performance/src/benchmarks/micro/libraries/System.IO.Compression/CompressionStreamPerfTestBase.cs

Lines 11 to 28 in ccbeb4e

	public class Gzip : CompressionStreamPerfTestBase
	{
	public override Stream CreateStream(Stream stream, CompressionMode mode) => new GZipStream(stream, mode);
	public override Stream CreateStream(Stream stream, CompressionLevel level) => new GZipStream(stream, level);
	}
	public class Deflate : CompressionStreamPerfTestBase
	{
	public override Stream CreateStream(Stream stream, CompressionMode mode) => new DeflateStream(stream, mode);
	public override Stream CreateStream(Stream stream, CompressionLevel level) => new DeflateStream(stream, level);
	}
	#if NET6_0_OR_GREATER // API introduced in .NET 6
	public class ZLib : CompressionStreamPerfTestBase
	{
	public override Stream CreateStream(Stream stream, CompressionMode mode) => new ZLibStream(stream, mode);
	public override Stream CreateStream(Stream stream, CompressionLevel level) => new ZLibStream(stream, level);
	}

I think it'd also be cleaner if both CreateStream overloads always set leaveOpen: true, so it doesn't differ by which is invoked in the tests.