First phase of the two-mode P3 async lifting work tracked under #94. This
PR lands the host-intrinsic ABI surface a stackful lift needs; the actual
trampoline emitter in adapter/fact.rs lands in the follow-up, mirroring
the #124 -> #129 split that worked for the data-plane intrinsics.

What lands:
- 4 new HostIntrinsic variants (ThreadNew, ThreadSwitchTo, ThreadYield,
  ThreadExit) with pinned signatures and a thread::* name module
- uses_stackful_lift() derived helper on P3AsyncFeatures
  (async lift without callback option => stackful mode per Component
  Model spec)
- Two new tests: stackful_intrinsic_signatures_pinned (4 names + sigs +
  pulseengine:async module) and stackful_lift_is_async_without_callback
  (4-case truth table)
- ADR-1 addendum documenting the two-mode policy side-by-side
- SR-32 flipped planned -> in-progress in safety-requirements.yaml and
  traceability.yaml updated to point at the new code + tests
- CHANGELOG [Unreleased] entry

Emitter behavior is unchanged in this PR: a component requesting
stackful lift still hits the existing clear error path. Phase 2 wires
the trampoline.

Refs: #94 (P3 async umbrella), SR-32 (stackful lifting requirement)
Milestone: v0.8.0

Co-authored-by: Claude Opus 4.7 <[email protected]>

…173)

* feat(p3): cross-component stream<T> pairing detection (#141, ADR-3)

Foundation for the in-module cross-component stream adapter — sub-A
of the P3 async umbrella #94.

Today meld lowers every component's stream operations to host
imports under `pulseengine:async` (ADR-1). When two fused components
share a `stream<T>` end-to-end — one holds the writable end, one the
readable end — both sides still route every chunk through the host
even though both ends now live in the merged module.

This PR ships the *detection* half (ADR-3, Path N): everything that
can be proven correct with unit tests and no wasm runtime. The
adapter *emitter* (ring buffer for same-memory, stream_read→copy→
stream_write chain for cross-memory) is a deliberately separate,
runtime-verified follow-up — cross-component stream codegen is
data-plane wasm that is only correct once executed on kiln/wasmtime,
and landing unverified codegen into the merged module is the H-1/H-3
hazard class meld's Mythos and LS-N gates exist to prevent. The repo
shipped P3 lowering in exactly this staged shape (#124 foundation,
#129 full pass).

What lands:

- `safety/adr/ADR-3-cross-component-stream-adapter.md` — design doc.
  Design paths: N (detection-now/emitter-next, chosen), O (full
  emitter in one PR, rejected — unverifiable), P (keep host routing,
  rejected — defeats fusion). Documents the two emitter shapes the
  follow-up implements and the foundation's precision boundary.

- `meld-core/src/p3_stream.rs` — new module, sibling to p3_async.rs.
  `StreamElement` (element type parsed from the `stream<...>`
  descriptor), `StreamRole` (Producer/Consumer from StreamWrite/
  StreamRead canonical entries), `StreamPair`, `StreamPairGraph`, and
  a pure `build_stream_pair_graph(components, resolved_imports,
  mode)`. The pairing logic is a pure function unit-tested without
  any ParsedComponent construction.

- `DependencyGraph` gains a `stream_pair_graph` field, populated by
  the resolver right after the resource graph (both key off
  `resolved_imports`). Memory mode follows `MemoryStrategy`.

Detection is conservative. A `StreamPair` is a *candidate*, recorded
only when two fusion-connected components have complementary roles
on a stream of the **same element type**. It does not prove the two
endpoints carry the same runtime handle (unknowable at build time).
Pairing only on matching element type — never any producer with any
consumer — keeps the foundation from manufacturing hallucinated
cross-type pairs ("hallucinations are more expensive than silence").

`meld-core/src/p3_stream.rs` added to the Mythos Tier-5 path lists in
mythos-gate.yml and mythos-auto.yml, per the documented drift-review
process for new fusion-correctness files.

Tests: 9 unit tests in p3_stream.rs including the 4 ADR-3 gating
fixtures. Full meld-core lib suite 239 → 247 passing. Clippy clean.

Refs: #141, #94 (umbrella), ADR-3, SR-33 (detection half).

Co-Authored-By: Claude Opus 4.7 <[email protected]>

* fix(ci): drop Tier-5 path-list change from this PR

claude-code-action self-validates that the workflow invoking it has
content identical to the version on `main` — a security measure that
stops a PR from altering the workflow to exfiltrate secrets. PR #173
modified `mythos-auto.yml` (adding `p3_stream.rs` to the Tier-5 path
list), so the action refused to run with:

  Action failed with error: Workflow validation failed. The workflow
  file must exist and have identical content to the version on the
  repository's default branch.

Reverts the `mythos-gate.yml` / `mythos-auto.yml` path-list edits so
`mythos-auto.yml` on this branch matches `main` and the auto-runner
can actually scan the Tier-5 source this PR touches (`resolver.rs`,
`merger.rs`). Registering `p3_stream.rs` in the Tier-5 lists moves to
a separate follow-up PR that touches no Tier-5 *source* — its
`detect` job sees no Tier-5 source change, sets `any=false`, and the
scan (hence the self-validation) never runs.

Co-Authored-By: Claude Opus 4.7 <[email protected]>

---------

Co-authored-by: Claude Opus 4.7 <[email protected]>