chore(splinter): initial commit

[psteinroe]

adds a new pgls_splinter crate with the basic integration.

its not hooked up yet.

pr includes

• the query execution
• the transformation into splinter diagnostics
• a codegen tool for the diagnostics categories that extracts them from the sql file
• a build file that downloads and parses the splinter.sql file from the repo. its cached and committed. The build script is just a helper when we want to "upgrade". we only have to change the SHA in build.rs for that.
• a check that splinter is only executed if supabase roles are in the target database

the query currently crashes when executed against a non-supabase database because of the calls to pg_catalog.has_table_privilege('anon', c.oid, 'SELECT') in some of the rules. for now, we skip splinter if the roles are not there. I opened an issue to ask for permission to add a flag to each rule so it can be used outside of supabase too: supabase/splinter#135. I would rather not make it our own just yet, but thats certainly a simple fix if things take too long.

UPDATE: manually split up the query into supabase-only and generic rules in a follow-up.

follow up prs will include:

• refactoring of the analyser codegen so we can reuse it for other lint tools to generate config, docs etc
• integration of the new DatabaseLocation attribute of the diagnostics struct (pr is open)