Tree-sitter: Split up ast_node_info table into two tables

[hvitved]

The DIL generated for AstNode.getParent is unnecessarily complex. Instead of using the ast_node_info extensional directly, we get a join with

TreeSitter::Ruby::AstNode#941e5e6b#b(
  /* TreeSitter::Ruby::AstNode.extends */ interned unique entity this
)
{
  (
    exists(
      /* @ruby_underscore_method_name */ interned dontcare entity _,
      /* @ruby_underscore_method_name */ interned dontcare entity _1
     |
      ruby_alias_def(this, _, _1)
    ) and
    project#ruby_ast_node_info(this)
  )
  or
  (ruby_alternative_pattern_def(this) and project#ruby_ast_node_info(this))
  // or
  // ... one case for each member @ruby_ast_node

This happens because ast_node_info needs to contain a row for each AST node, as it also records AST node locations, and hence the column for the parent is not simply an AstNode, but either an AstNode or a File (to account for root AST nodes). By splitting up ast_node_info into two separate extensionals, ast_node_location and ast_node_parent, we can use the ast_node_parent extensional directly in AstNode.getParent.

The decrease in DIL size is confirmed by DCA.

[nickrolfe]

Interesting. I wonder why the original ruby_ast_node_parent table (before #7875 renamed it and added the location column) could represent file-parents if we never relied on it in QL.

Anyway, the changes LGTM.