[2] refactor: use more specific tree-sitter references and identifiers

[juleswritescode]

• introduces a couple of new leaf-identifier nodes: role_identifier, table_identifier and so on
• introduces a couple of new referenves next to the old object_reference, e.g. table_reference or function_reference.
• adds queryable tags to the fields of the references and refactors the tree-sitter queries to use them

Notice that we can only use precise leaf-identifiers if we have a fully qualified *_reference:

In an incomplete statement, if a user writes select us| from some_table;, the us could be either a schema, a table, an alias, or a column. The user is still writing the query; we don't know what he'll type next!

Even if he writes select something.us|, something might refer to a table, alias or schema; us might refer to a table or column.

Only when he writes select something.users.email do we know which part is a schema, which a table, and which a column.

The tags on the *_reference should help us decide which data types to consider for completion/hover.

---

In the next PRs, I want to store relevant parent nodes and field names. Right now, we're traversing the tree alot, for example to find out if a node has a certain tag or if the ancestory history matches a template. We can make this easier and faster by collecting the traversed nodes/fields in some kind of data structure and checking against that.