fix(swift): index protocol method requirements as Method nodes

[DeusData]

fix(swift): index protocol method requirements as Method nodes

Advances #43. Distilled from #613 by @xbsjason, credited via Co-authored-by.

The gap

A Swift protocol requirement — a bodyless func generate() -> String inside a protocol — produced no node at all. Swift codebases are heavily protocol-driven, so the requirement is frequently the declaration a reader actually wants, and "who declares generate()" simply had no answer.

What delivers it

Three edits, and I verified each is load-bearing by removing it and watching the test fail:

1. protocol_body joins the class-body types in find_class_body — without it, a protocol's members are never walked.
2. protocol_function_declaration joins swift_func_types — extract_class_methods gates on that set, so without the entry the requirement gets walked and then discarded.
3. Both name resolvers accept the node. It has the same simple_identifier shape as function_declaration, and since it now reaches the free-function path too, naming it in only one place would leave a latent unnamed-def bug.

What I dropped from #613, and why

The enum/struct half. #613 also tried to make bare struct/enum produce Class/Enum nodes, and both of its assertions fail on CI (lsan-macos, test-unix, and Windows alike). The cause is worth recording: tree-sitter-swift has no struct_declaration or enum_declaration node type — it models both as class_declaration:

class_declaration:             2
protocol_declaration:          1
protocol_function_declaration: 1
protocol_body:                 1
struct_declaration:            0   ← not a symbol
enum_declaration:              0   ← not a symbol

So those two entries in swift_class_types have been inert since they were written, class_label_for_kind's enum_declaration → "Enum" branch never fires for Swift, and a bare Swift enum is labeled Class. That is a genuine pre-existing modelling gap. I've left the dead entries in place with a comment naming the problem — deleting them would tidy the code while erasing the only marker that the gap exists — and it stays out of this PR.

A find_first_descendant_by_kind rescue in extract_class_methods. Removing it changes no test outcome once the function-type entry is present, so it was carrying nothing.

The result is three production edits instead of six, with no behaviour left unproven.

Tests

swift_protocol asserts Interface:StudyRunning + Method:generate. It is binding: applied against unmodified main, it fails on exactly ASSERT(has_def(r, "Method", "generate")).

Verification

macOS ASan/UBSan: extraction, grammar_labels, grammar_probe_g, grammar_regression, lang_contract, parse_coverage — 384 passed, 0 failed. lint-ci green.