docs: the Darwin SO_NOSIGPIPE rule is written twice under sendNoSig and sendNoSigNoWait

open   by a73x

docs/decisions.md describes the Darwin SO_NOSIGPIPE rule twice under two names and never says they are one rule shared by a pair.

The platform-layer entry (around 'sendNoSigNoWait sets SO_NOSIGPIPE, and a REFUSED set is the answer') names the SERVER arm `server_os.sendNoSigNoWait`; the Darwin-arm entry (around 'sendNoSig sets SO_NOSIGPIPE per call') names the CLIENT arm `client_os.sendNoSig`. Both are correct for the function they name, but a reader meeting one entry cannot tell whether the other function has the rule too, or whether the name drifted between the two dates. In the tree both Darwin arms (src/os/client_os_macos.zig, src/os/server_os_macos.zig) set the option per call, read EINVAL as BrokenPipe, and send through `sendto` mapping ENOTCONN to BrokenPipe; the switch is duplicated verbatim because a shared file would need a third module row.

Wanted: one sentence in each entry saying the pair shares the rule, and a cross-reference between them. Docs only; no code change. Noted by the branch's final re-review on 2026-09-04 and left out of scope there.