fix(abi-verify): tolerate non-canonical Zig switch arm false shorthand (#36)

`abi-verify`'s Zig FFI parser bombed on 5 cartridges' `isValidTransition`
switch arms because their terminal-state arm body is the literal
`false` (no outgoing transitions allowed) rather than the canonical
`to == .<v>` chunk form:

    fn isValidTransition(from: BspState, to: BspState) bool {
        return switch (from) {
            ...
            .exited => false,   // <-- parser bombed here
        };
    }

The Zig is well-formed and the semantics ("empty allowed-set") are
clear; the verifier just didn't accept the shorthand.

`parse_arm_targets` now detects this form (body trimmed of trailing
`,`/`;` and whitespace equals `false`) and returns the empty vec —
equivalent to the "this state has no allowed outgoing transitions"
manifest semantics, which is exactly what the cartridges intend.

End-to-end verified against the 5 cartridges named in the issue
(after a fresh `cargo build --release`):

  bsp-mcp        → parses cleanly; surfaces real drift on BspCapability
  container-mcp  → abi-verify OK
  dap-mcp        → parses cleanly; surfaces real drift on StepGranularity
  lsp-mcp        → parses cleanly; surfaces real drift on CompletionKind
  vault-mcp      → parses cleanly; surfaces real drift on IdentityType

All 5 now produce either exit 0 (clean) or a real drift diagnosis,
which is precisely the acceptance criterion in iseriser#19. The
post-parse drift findings are separate per-cartridge issues — not
verifier defects — and out of scope for this PR.

44 lib tests + 9 integration tests pass.

Refs hyperpolymath/standards#92 (Phase 2 allowlist expansion).
Refs #19.

Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

Author: hyperpolymath