Add support for semantic predicates

[ydah]

Overview

This PR adds support for semantic predicates to Lrama, allowing conditional enabling of grammar rules based on runtime conditions.

Syntax:

rule : {expression}? TOKEN { action } | TOKEN { action } ;

The predicate {expression}? is evaluated at parse time. If it returns true (non-zero), the alternative is enabled.

Example:

widget : {new_syntax}? WIDGET ID NEW_ARG { printf("New syntax\n"); } | {!new_syntax}? WIDGET ID OLD_ARG { printf("Old syntax\n"); } ;

Motivation / Background

Semantic predicates enable context-sensitive parsing, which is useful for:

• Version-dependent syntax (e.g., supporting both old and new language features)
• Context-sensitive keywords (e.g., async in JavaScript that behaves differently based on context)
• Conditional grammar rules based on parser state

This feature is similar to ANTLR4's semantic predicates and fills a gap in Lrama's capabilities for handling context-dependent grammars. Leading predicates (at the start of a rule) affect prediction, while trailing predicates act as validation.