Make the per-grammar regression suite fully green. (1) Add def/name extraction for 22 IDL/config/niche grammars that previously yielded only a Module node (agda, pony, move, cobol, janet, pine, smali, verilog, vhdl, systemverilog, protobuf, graphql, thrift, capnp, smithy, wit, prisma, cmake, puppet, tablegen, assembly, nasm): corrected function/class node-type sets in lang_specs.c and per-grammar name resolution in extract_defs.c (these grammars define no tree-sitter 'name' field, so names are read from typed child nodes; a depth-bounded descendant search handles the FIELD_COUNT-0 (System)Verilog wrappers). (2) Resolve same-file CALLS for groovy, solidity, and ada (callee handlers in extract_calls.c plus descending into Ada subprograms). Golden label histograms and fixtures are updated to the real new extraction output (min_defs raised, never lowered).
Martin Vogel
d30093aa8f