chore(base-parsing, base-expression, base-template-processor): wire Rule into ExpressionParser/TemplateParser and port ELParser/JtParser to AbstractParser

[deer]

• ExpressionParser<N> and TemplateParser<N> now implement Rule<N> / Rule<List<N>>, exposing a tryParse(Scanner) entry point that stops on unrecognised tokens rather than throwing — backed by the new Tokenizer.tokenize(Scanner) overload. The existing parse(String) methods are preserved and delegate to tryParse.
• ELParser is ported to AbstractParser<Node>: the manual Scanner field, followsKeyword/consumeKeyword private methods, and hand-rolled error translation are all removed in favour of inherited infrastructure. parseComposite replaces its manual depth-tracking loop with Scanner#consumeBalanced.
• JtParser.parse now accepts a String (full file content) instead of List<String>; the separate HeaderParser is removed and replaced by a new JtFileParser extends AbstractParser<ParsedTemplate> that handles the entire file in one scanner pass. TemplateProcessor uses Files.readString at the call site.