Add Product recommendation sample that will suggest k best related products to subject product

Provided sample it's not quite useful in real word scenarios. As basically the problem stated as pick k best-suited products to the subject product. Using the current approach I should recalculate scoring for all universe of products from N total products. So to form recommendations for whole store I should recalculate n*n scores and sort to pick the best recommendation, which looks too compute intensive even if this is quite parallelizable computations for several thousand products it's quite expensive.

As per @wschin suggestion, we can try and implement

> There are some approximated solutions (**approximated maximum inner product search and approximated nearest neighbor search usually via K-D tree**) but ML.NET doesn't support any of them.