chore(beads): ftw/nftw ACTIONRETVAL findings + repo build-break triage (bd-nwrdyu, bd-650gx2, bd-3g3u9z, bd-aaouxv, bd-ggef4a, bd-78f8do, bd-6rrgjs)

Tracker state for this session's work:
  bd-650gx2 CLOSED  workspace manifest would not load (harness fixture-exec override)
  bd-3g3u9z CLOSED  frankenlibc-core would not compile (strspn_set/strcspn_set)
  bd-aaouxv OPEN    frankenlibc-abi still will not compile; ~20 committed errors from
                    merge 0be7b9d, split into a mechanical half (restore definitions
                    from parent 1c8f9f7) and a structural half (ScanSpec.alloc,
                    StepOutcome::LongRoute) that their owning campaigns must resolve
  bd-nwrdyu OPEN    reopened and blocked on bd-aaouxv; the title's premise was wrong
                    (ACTIONRETVAL was implemented, its glibc parity was just unproven)
  bd-ggef4a NEW     nftw dangling-symlink flags inverted vs glibc in both walk modes
  bd-78f8do NEW     nftw logical walk lacks (dev,ino) revisit suppression -> unbounded
                    recursion on symlink cycles
  bd-6rrgjs         measured live-glibc gethostbyname hosts-merge semantics attached as
                    a comment for the agent already working it; not claimed

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Author: Dicklesworthstone