[cpp] fix Semgrep pattern parsing for class/struct/enum bodies and operator overloads

[brandonspark]

Summary

Generated release artifacts from semgrep/ocaml-tree-sitter-semgrep#581.

Augments the C++ grammar so canonical Semgrep patterns parse cleanly:

• LANG-486: ... and $X inside class/struct bodies (via _field_declaration_list_item) and enum bodies (via a wholesale override of enumerator_list).
• LANG-497: operator $OP as an operator name (via a previous-style augmentation of operator_name).

The semgrep_metavar token gets a precedence bump so the lexer prefers it over identifier (whose regex also accepts a leading $) when both are valid at a position; this lets a bare $X parse as a metavariable inside enum/struct bodies instead of as a malformed declaration or enumerator.

Upstream change: semgrep/ocaml-tree-sitter-semgrep#581

Test plan

• Generated by ./release --branch fix/cpp-pattern-augmentation cpp from the upstream branch.
• All 201 corpus tests pass on the upstream side, including 6 new tests for the new patterns.

Generated with Claude Code