fix(transport)(sphere-sdk#559): opt out HMCP + ACP sendDM self-wraps

[vrogojin]

Summary

F3 from the #559 verdict comment. Wires { selfWrap: false } (new SDK API from sphere-sdk PR #558) into the two short-lived RPC paths the F1 soak identified as the biggest contributors to §3/§4 relay-index pollution:

• src/transport/dm-transport.ts (HMCP — CLI → host manager)
• src/trader/acp-transport.ts (ACP — CLI → trader tenant)

src/legacy/legacy-cli.ts:3961 (sphere message send) is not touched — that command is user-initiated and the user likely wants their sent DM in their own outbound history. Default behavior preserved.

Why

Both files implement short-lived RPC: the sphere host … / sphere trader … process publishes a single request, awaits the correlated reply, and exits. The default NIP-17 self-wrap publishes a SECOND gift-wrap targeting the SENDER's own pubkey so a returning client can replay outbound history — useless work for one-shot CLI, and exactly the noise the F1 soak measured at ~7 new events per spawn process in §3 of manual-test-trader-roundtrip.sh. Cutting the self-wrap halves the per-process publish count and removes the corresponding entries from the relay's #p:controller index for the next short-lived CLI process to pull down on subscribe.

The fix is a strict subset of what sphere-sdk PR #558 was designed for — see #558's own description: "For short-lived senders (one-shot CLI RPC like sphere trader portfolio) the process exits before it could ever read its own self-wrap, making the second publish pure relay-index pollution."

What changed

Both dm-transport.ts and acp-transport.ts:

• Widen the local SphereComms.sendDM interface to forward an optional SendMessageOptions to the SDK call.
• Pass { selfWrap: false } at the only outgoing sendDM call site.

Net diff: 2 files changed, 24 insertions(+), 6 deletions(-).

Test plan

• npx tsc --noEmit → no new errors (3 pre-existing errors in legacy-cli.ts confirmed by stashing F3 and re-running)
• npx vitest run --reporter=basic → 187/187 pass (incl. 14 dm-transport, 19 spawn, 44 trader-commands)
• (Follow-up) Re-run F1 soak with F3 + sphere-sdk #558 both landed; expect per-spawn new-event count to drop from ~7 to ~3-4 (HM responses remain, CLI self-wraps gone)

Related

• sphere-sdk#555 — original M8 mitigation framing
• sphere-sdk#559 — fresh-wallet backlog + F1 soak verdict
• sphere-sdk PR #558 — selfWrap: false SDK API (merged into uxf)
• sphere-sdk PR #582 — F1 instrumentation (open, awaiting follow-up soak)
• sphere-sdk PR #614 — re-export SendMessageOptions from index.ts (merged, prereq for this PR's typecheck)