Add llms evals which measures how well llms use our docs (not llms.txt only) to create applications

[cloutiertyler]

[bfops]

We need to spec this out a bit more (DoD etc)