RLVR expansion

[natolambert]

Will expand PR iteratively until done and then merge and push to Arxiv.
Currently:

• Expand "training overview" chapter to include summaries of canonical training recipes.
• Add summary of async rl infrastructure
• More reasoning related works (pre o1)
• Summary of major reasoning reports to date
• Summary of methods used across reasoning efforts to see trends
• Some other minor additions and fixes throughout