Import tensors and add methods to sbox

[ArtisticRoomba]

CC @moonheart08

This PR adds System.Numerics.Tensors to the RT project and adds some methods to the sandbox.

In general, Tensors is a much better way to take advantage of SIMD intrinsics on hardware without coding all of the methods ourselves (I felt really silly adding AVX512 impls to all of our current NumericsHelpers methods).

Useful for Atmospherics in SS14.

Microbenchmarks shows improvement, but of course it doesn't paint a full picture:

Details

BenchmarkDotNet v0.15.8, Linux Gentoo Linux
AMD Ryzen 9 7900X 4.03GHz, 1 CPU, 24 logical and 12 physical cores
.NET SDK 10.0.101
  [Host]     : .NET 10.0.1 (10.0.1, 10.0.125.57005), X64 RyuJIT x86-64-v4
  Job-JMDAGQ : .NET 10.0.1 (10.0.1, 10.0.125.57005), X64 RyuJIT x86-64-v4

Toolchain=.NET 10.0  

Method	N	Mean	Error	StdDev	Code Size	Allocated
BenchNumericsHelpers	32	3.110 ns	0.0624 ns	0.0521 ns	637 B	-
BenchNumericsHelpers	128	10.249 ns	0.0312 ns	0.0244 ns	637 B	-
BenchNumericsHelpers	256	12.450 ns	0.1919 ns	0.1701 ns	637 B	-
BenchNumericsHelpers	512	25.505 ns	0.3755 ns	0.3513 ns	637 B	-
BenchNumericsHelpers	1024	53.923 ns	0.8464 ns	0.7918 ns	637 B	-
BenchNumericsHelpers	2048	90.217 ns	0.9333 ns	0.8273 ns	637 B	-
BenchNumericsHelpers	4096	594.813 ns	5.8739 ns	5.4944 ns	637 B	-
BenchNumericsHelpers	8192	623.864 ns	6.7247 ns	6.2903 ns	676 B	-
BenchNumericsHelpers	16384	1,300.781 ns	9.5156 ns	8.4353 ns	676 B	-
BenchTensor	32	3.574 ns	0.0217 ns	0.0193 ns	1,367 B	-
BenchTensor	128	9.457 ns	0.1164 ns	0.1089 ns	1,374 B	-
BenchTensor	256	15.379 ns	0.1173 ns	0.1040 ns	1,369 B	-
BenchTensor	512	26.599 ns	0.1916 ns	0.1792 ns	1,369 B	-
BenchTensor	1024	47.837 ns	0.1911 ns	0.1596 ns	1,369 B	-
BenchTensor	2048	91.371 ns	0.7019 ns	0.6222 ns	1,369 B	-
BenchTensor	4096	304.175 ns	1.8590 ns	1.7389 ns	1,369 B	-
BenchTensor	8192	652.692 ns	5.9393 ns	5.5556 ns	1,369 B	-
BenchTensor	16384	1,198.696 ns	16.3505 ns	15.2943 ns	1,369 B	-

And the emitted assembly seems to show that AVX-512 instructions are being used in BenchTensor compared to AVX-256 instructions in BenchNumericsHelpers.

[moonheart08]

These are from Systems.Buffers, while they're in the numerics tensors assembly this works based on namespace.

Additionally, no consuming APIs are provided, so these shouldn't be either.

[moonheart08]

No APIs are provided that can consume this type, so it shouldn't be provided.

[ArtisticRoomba]

Without it, attempting to create a simple tensor using Tensor.Create(); will cause a sandbox violation:

Access to type not allowed: [System.Numerics.Tensors]System.Numerics.Tensors.Tensor`1

[moonheart08]

Without it, attempting to create a simple tensor using Tensor.Create(); will cause a sandbox violation:

Access to type not allowed: [System.Numerics.Tensors]System.Numerics.Tensors.Tensor`1

Okay but you can't use the Tensor, so why allow Tensor.Create either?

[ArtisticRoomba]

Ah, right, I see what you mean now. Genius me.

[moonheart08]

never used.