Model the query dependency graph

(to be done after the salsa migration)
No one really has a clear picture on our queries, their individual purposes, their dependencies, etc. Given we are going to touch all of this with the salsa migration over time we might as well try to document all of this properly (and maybe generate a fancy visual dependency graph while we are at it).

What's the use in this? The dependencies we draw between queries are fairly important wrt to incrementality, which I believe has severely worsened the past 2 years, causing us to become fairly slow for (supposedly) trivial edits in documents! (one reason for this is the span map, it is not incremental friendly... we likely need to add an extra firewall query for that).

And on the topic of the word firewall query, a glossary might be nice as well :)

This is intended for after the salsa migration. Start by reviewing the query system and the span map, then map each query's purpose and dependencies, including where a firewall query may be needed. Done means the query dependency structure is documented, with a decision on whether to add a visual graph and glossary.