Skip node locating when the document is not parsed

[st0012]

Motivation

Although DocumentHighlight#run skips #visit when the document is not parsed, DocumentHighlight#initialize doesn't skip node locating.

This means that DocumentHighlight will fail to initialize and cause error when the source code has syntax error.

Fixes #326

Implementation

Check if the document is parsed before trying to find the target node from document.tree.

Automated Tests

Added a test to make sure the DocumentHighlight request can still be initialized with syntax-incorrect source code.

Manual Tests

1. Create a file that has syntax error, like

   def foo(bar
     bar
   end
2. Switch to OUTPUT tab
3. Move cursor on foo to trigger documentHighlight requests
4. • On main, you'd see

     Received response 'textDocument/documentHighlight - (40)' in 2ms. Request failed: #<NoMethodError: undefined method `child_nodes' for nil:NilClass
     
             matched = parent.child_nodes.compact.bsearch do |child|
                             ^^^^^^^^^^^^> (-32603).
     
   • On this branch, there should be no error