Add a readme under each benchmark.

[Anindyadeep]

Under the docs section we have a information about ML-Engines and benchmarking information for Llama2 models in Llama2-template. However, we do not provide information about each of the benchmark. So, this issue proposes to provide a readme under benchmark folder. Under the readme the following information should be provided:

• A little desc about of the benchmark.
• A original links of the implementation.
• device supported.
• quantization.
• The limitations of the benchmark.
• How to run that particular benchmark on custom data example.

This will help newcomers to easily get started with a new framework or a backend and use it properly.

[Anindyadeep]

This is the updated structure:

• About (This includes)

  • general desc
  • paper (if any)
  • implementation source

• Benchmarks info

  • The numerical info
  • Any specific information like supported devices and limitations (can also reference the issues related to it)