perf(actions): pre-normalize Command Palette fields and eliminate repeated lowercasing in schema search #1528

Problem

1. In QueryaCommand.matches (lib/core/actions/querya_command.dart), for every registered command on every keystroke in the Command Palette, id.split('.').last.toLowerCase(), fields = <String>[...], and q.split(RegExp(r'\s+')) are executed. For 150 commands, this creates 150 regex compilations and hundreds of temporary lists and strings.
2. In filterSchemaObjects (lib/core/actions/querya_schema_object.dart), for every keystroke, object.searchHaystack.toLowerCase() is called for all 5,000+ schema objects in memory, generating 5,000 heap strings per character typed.

Solution

• Pre-normalize search haystacks to lowercase once when creating or registering QueryaCommand and QueryaSchemaObject.
• Split query once before the loop rather than repeating inside every command's match check.
• Replace id.split('.').last with index scanning from the end (lastIndexOf('.')).

Benefits

• 0 lowercasing allocations inside the search loop (saving 5,000+ string allocations per keystroke).
• Smooth 60/120 FPS filtering in Quick Open and Command Palette.