Rework of metrics and documentation for Omniperf profiling

[skyreflectedinmirrors]

This PR contains the collective work of many AMDer's (including myself, @gsitaram, @IanBogle, @rajaroracmu, @ausellis0, @coleramos425, @rwvo, @jlgreathouse and others) that:

• Reworks Omniperf metrics for existing CDNA accelerators (e.g., MI250) for improved user experience / usability, etc.
• Adds full documentation of the updated metrics
• Adds several examples, and discussion of using Omniperf on said examples

[dmcdougall]

@skyreflectedinmirrors Are you looking for feedback on this yet?

[skyreflectedinmirrors]

For sure, any feedback / corrections are welcome. I will note that:

• A) this isn't intended to be a profiling guide, more "what do these mean"
• B) the scope doesn't include MI300's (yet)

Feel free to reach out to me internally as well for any specific questions that might not be suitable for a public discussion

[coleramos425]

I suggest modifying copyrights in the new sample code to include our delimiters. This way things are compatible with utils/update_license.py