this rewires the workspace server for better DX and adds support for sql function bodies.

closes#178

hear me out:

• theDocument is responsible for extracting statements from a source file. right now, this is always SQL, but in the future we could make this generic.
• theParser is the new layer between the workspace server and the document. there is one parser per document, and the parser owns the document.
• the parser is responsible for parsing the raw sql coming from the document, and is the main API layer for the workspace.
• the consumer of the parser apis simply gets the statements and the related parse results. the consumer does not need to know anything about sub statements
• I added a nice pattern where the consumer of the Parser can define a mapper and / or a filter and then simply callsparser.iter(ExecuteStatementMapper, CursorPositionFilter::new(pos)) and gets exactly the data it needs.
• I also optimised the parsing in general: the only thing that actually cares about modified statements is tree sitter. for pg_query, we always just remove and add when we receive a modify statement. hence, we can simply lazily parse all but tree sitter and just use modified / remove notifications to clear the cache. this also opened up the way for a nicer api because if you want the pg_query results, you always get Result<Node, SyntaxError> and we do not need to wrap it in an Option<> anymore.
• the statement id is now not just an usize anymore but an enum that can reflect sub statements. if we remove parse results for a parent, we also remove it for its sub statement.