Streams

[mgravell]

Adds a new SequenceStream / ReadOnlySequenceStream pair - similar to MemoryStream, but:

• GetBuffer() returns [ReadOnly]Sequence<byte>
• when created as dynamically resizeable, it leases segments from the ArrayPool<byte>.Shared and returns them when trimmed/disposed

Benefits:

• buffer allocations are pooled leases
• because non-contiguous memory is the norm, there's no need to copy any data around when expanding the underlying buffer (in this use-case, SequenceStream owns the sequence, so it can play with the chain; it does not mutate the chain if an existing one is passed in)
• allows you to treat a Sequence<byte> or ReadOnlySequence<byte> as a Stream with very low overhead
• (marginally) out-performs MemoryStream anyway (I'd have settled for "equal or slightly worse")

Possible additions:

• should we allow a specific ArrayPool<byte> to be specified? or is that unnecessary?
• if so, should we actually be exposing/using MemoryPool<byte>[.Shared] instead of ArrayPool<byte>?
• should we allow arena-based allocation as an optional opt-in?

Performance:

BenchmarkDotNet=v0.11.4, OS=Windows 10.0.17134.648 (1803/April2018Update/Redstone4)
Intel Core i7-5930K CPU 3.50GHz (Broadwell), 1 CPU, 12 logical and 6 physical cores
Frequency=3416973 Hz, Resolution=292.6567 ns, Timer=TSC
  [Host]     : .NET Framework 4.7.2 (CLR 4.0.30319.42000), 64bit RyuJIT-v4.7.3362.0
  Job-FQKGZY : .NET Framework 4.7.2 (CLR 4.0.30319.42000), 64bit RyuJIT-v4.7.3362.0
  Job-UGUEYK : .NET Core 2.2.2 (CoreCLR 4.6.27317.07, CoreFX 4.6.27318.02), 64bit RyuJIT

Method	Runtime	Toolchain	Mean	Error	StdDev	Median	Ratio	RatioSD	Gen 0/1k Op	Gen 1/1k Op	Gen 2/1k Op	Allocated Memory/Op
MemoryStreamDefault	Clr	net472	8.933 ms	0.1760 ms	0.3515 ms	9.013 ms	1.07	0.05	390.6250	390.6250	390.6250	1399112 B
MemoryStreamPreSize	Clr	net472	8.263 ms	0.1633 ms	0.3480 ms	8.466 ms	1.00	0.04	140.6250	140.6250	140.6250	515976 B
RecyclableMemoryStreamDefault	Clr	net472	7.570 ms	0.0628 ms	0.0491 ms	7.552 ms	0.92	0.01	-	-	-	1024 B
RecyclableMemoryStreamPreSize	Clr	net472	7.899 ms	0.1538 ms	0.2206 ms	7.903 ms	0.95	0.03	-	-	-	960 B
SequenceStreamDefault	Clr	net472	7.863 ms	0.1568 ms	0.1743 ms	7.797 ms	0.96	0.02	-	-	-	448 B
SequenceStreamPreSize	Clr	net472	8.606 ms	0.2415 ms	0.7120 ms	8.407 ms	0.98	0.04	-	-	-	512 B
MemoryStreamDefault	Core	netcoreapp2.2	8.234 ms	0.1316 ms	0.1167 ms	8.193 ms	1.00	0.00	390.6250	390.6250	390.6250	1397576 B
MemoryStreamPreSize	Core	netcoreapp2.2	8.250 ms	0.1647 ms	0.2927 ms	8.199 ms	1.01	0.04	140.6250	140.6250	140.6250	515048 B
RecyclableMemoryStreamDefault	Core	netcoreapp2.2	7.999 ms	0.1564 ms	0.2389 ms	8.036 ms	0.97	0.04	-	-	-	920 B
RecyclableMemoryStreamPreSize	Core	netcoreapp2.2	7.888 ms	0.1541 ms	0.2575 ms	7.861 ms	0.97	0.03	-	-	-	920 B
SequenceStreamDefault	Core	netcoreapp2.2	8.282 ms	0.1627 ms	0.3359 ms	8.334 ms	1.00	0.05	-	-	-	400 B
SequenceStreamPreSize	Core	netcoreapp2.2	8.227 ms	0.1629 ms	0.3364 ms	8.253 ms	1.01	0.03	-	-	-	400 B