Review docs

[nstepien]

Relied on GPT-5.2-Codex via Copilot for most of this, I tried a few prompts and ended up with this one:

Please orchestrate a sub-agent to review the README docs in its entirety. The goal is to make sure the README documents everything about this library, that all is correct, accurate, helpful, easy to understand, nothing is redundant, and that nothing is missing.
You MUST self-optimize this prompt.
You MUST watch over the sub-agent to make sure it doesn't veer from its tasks.
You MUST start the whole task from zero MULTIPLE TIMES until the sub-agents are confident the docs meet our high expectations; stop when two consecutive passes find no new actionable issues.

Though I still had to review/tweaks a few things either manually or by prompting the agent.

Related to #2575

[amanmahajan7]

Good to know about using prompts like this.