Prepare for v0.6 release

[fsschneider]

This PR prepares for the upcoming v0.6.0 release of the benchmark. Depends on #874.

Specifically, it changes:

• Rework Readme to be clearer and easier to understand.
• Remove deprecated COMPETITON_RULES and CALL_FOR_SUBMISSION docs.
• Backfill the changelog with the most relevant changes and highlight it in more documents.
• Increment rules version to 0.6.0 and detail new rule changes:
  • Reduce studies from 5 to 3.
  • Remove held-out workloads.
  • Update API in documentation.
• Integrate and describe the new version policy (Proposal: Unify Versioning for Codebase, Leaderboard, and Rules #870).
• Minor formatting improvements.

Todos

There are a few more rule changes to detail. I also want to give the entire DOCUMENTATION.md a thorough review, focusing on readability and removing deprecated statements. The changes include, for example:

• Remove mentions of model track?
• Rewrite section on qualification set -> Phrase it as a recommended protocol for reporting partial results if limited resources are available. An AlgoPerf light version?
• Update note on "how many submissions"
• Update FAQs (for rolling leaderboard format)
• Integrated important requirements from the COMPETITON_RULES doc (e.g., Apache 2.0 license, defining software dependencies, CLA/membership/trademark agreement?)
• Mention that you can change the dropout value in the forward pass now.

[github-actions]

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅