Update System.Numerics.Tensors PACKAGE.md to reflect current library capabilities (#118512)

* Initial plan

* Update System.Numerics.Tensors PACKAGE.md to reflect current library capabilities

Co-authored-by: stephentoub <[email protected]>

* Update src/libraries/System.Numerics.Tensors/src/PACKAGE.md

Co-authored-by: Tanner Gooding <[email protected]>

* Fix trailing spaces in PACKAGE.md markdown linting issue

Co-authored-by: stephentoub <[email protected]>

---------

Co-authored-by: copilot-swe-agent[bot] <[email protected]>
Co-authored-by: stephentoub <[email protected]>
Co-authored-by: Stephen Toub <[email protected]>
Co-authored-by: Tanner Gooding <[email protected]>

Author: 3 people

src/libraries/System.Numerics.Tensors/src/PACKAGE.md

@@ -1,12 +1,15 @@
 ## About
 
-Provides methods for performing mathematical operations over _tensors_ represented as spans.  These methods are accelerated to use SIMD (Single instruction, multiple data) operations supported by the CPU where available.
+Provides methods for performing mathematical operations over _tensors_. This library offers both high-level tensor types and low-level primitives for working with multi-dimensional numeric data.  Many operations are accelerated to use SIMD (Single instruction, multiple data) operations supported by the CPU where available.
 
 ## Key Features
 
-* Numerical operations on tensors represented as `ReadOnlySpan<float>`
+* High-level tensor types: `Tensor<T>`, `TensorSpan<T>`, `ReadOnlyTensorSpan<T>` for working with multi-dimensional arrays
+* Low-level tensor primitives: `TensorPrimitives` for efficient span-based operations
+* Generic support for various numeric types (float, double, int, etc.)
 * Element-wise arithmetic: Add, Subtract, Multiply, Divide, Exp, Log, Cosh, Tanh, etc.
 * Tensor arithmetic: CosineSimilarity, Distance, Dot, Normalize, Softmax, Sigmoid, etc.
+* SIMD-accelerated operations for improved performance
 
 ## How to Use
 
@@ -22,6 +25,7 @@ var movies = new[] {
 };
 var queryEmbedding = new[] { 0.12217915f, -0.034832448f };
 
+// Using TensorPrimitives for low-level span operations
 var top3MoviesTensorPrimitives =
     movies
         .Select(movie =>
@@ -36,13 +40,24 @@ foreach (var movie in top3MoviesTensorPrimitives)
 {
     Console.WriteLine(movie);
 }
+
+// Using higher-level Tensor types for multi-dimensional operations
+float[] data1 = [1f, 2f, 3f, 4f, 5f, 6f];
+float[] data2 = [6f, 5f, 4f, 3f, 2f, 1f];
+Tensor<float> tensor1 = Tensor.Create(data1, [2, 3]); // 2x3 tensor
+Tensor<float> tensor2 = Tensor.Create(data2, [2, 3]); // 2x3 tensor
+Tensor<float> result = tensor1 + tensor2;
 ```
 
 ## Main Types
 
 The main types provided by this library are:
 
-* `System.Numerics.Tensors.TensorPrimitives`
+* `System.Numerics.Tensors.TensorPrimitives` - Low-level operations on spans of numeric data
+* `System.Numerics.Tensors.Tensor<T>` - Generic tensor class for multi-dimensional arrays
+* `System.Numerics.Tensors.TensorSpan<T>` - Span-like view over tensor data
+* `System.Numerics.Tensors.ReadOnlyTensorSpan<T>` - Read-only span-like view over tensor data
+* `System.Numerics.Tensors.Tensor` - Static class with high-level tensor operations
 
 ## Additional Documentation