feat(modules/cargo): cargo/build + cargo/test — structured Rust toolchain wrappers

[joelteply]

Summary

Closes Priority 2 from PERSONA-AS-DEVELOPER-GAP.md: Rust iteration parity with TypeScript. Personas can now build + test their own scaffolded modules and get the same structured feedback density a human gets from npm run build:ts / cargo test.

What this adds

New stateless cargo ServiceModule at src/workers/continuum-core/src/modules/cargo/:

Command	Signature	Returns
cargo/build	{package?, features?, release?, workingDir?, timeoutMs?}	{success, errors: CargoMessage[], warnings: CargoMessage[], exitCode?, durationMs, error?}
cargo/test	{package?, filter?, features?, libOnly?, release?, workingDir?, timeoutMs?}	{success, passed, failed, ignored, measured, failures: string[], buildErrors: CargoMessage[], exitCode?, durationMs, error?}

Plus 6 ts-rs wire types: CargoBuildParams, CargoBuildResult, CargoTestParams, CargoTestResult, CargoMessage, CargoSpan.

Doctrine followed (per field manual)

• §3 Module Design Template — typed envelope at entry/exit, ts-rs annotations, camelCase serde, optional fields elided
• §4.1 Concurrency — stateless module (cargo handles its own target-dir locking; module-level lock unnecessary)
• §4.2 Concurrency tests — multi-thread tokio (worker_threads = 4) fires 8 parallel real-cargo subprocess invocations; every result consistent
• Three primitives — both commands are pure Commands. Streaming variants land when Stream cell shape ships (gap report priority 4)
• Rethink-not-port — designed Rust-first; no TS predecessor

Sharp design decisions (the kinks the tests caught pre-merge)

Bug	What broke	Fix
parse_summary_counts got 0 every time	Positional indices 0,1 broke on "ok. 22 passed" (token 0 is "ok.")	Scan WITHIN each chunk for first <int> <label> pair
Failures-block exit condition	Exited on lines containing : — but test names ARE module::test	Enter on failures:, capture single-token lines containing ::, exit on next test result:
Repeated failures: blocks	libtest emits TWO per failing binary (with stdout decorators + without)	Capture from both, dedupe by first-seen order, skip ---- ... ---- markers
Subprocess pipe deadlock potential	If child fills stdout buffer waiting for us	Concurrent tokio tasks for stdout/stderr read alongside wait()
Runaway timeout	Persona could hold substrate forever	Hard-cap: build 15min, test 30min (configurable lower)

Composability with the grid (the alignment payoff)

Per the gap report's "later parts of the vision" section: both result envelopes are flat camelCase JSON, trivially serializable across airc's grid. A persona on Joel's M-series Mac can call cargo/test against a module a persona on a peer's RTX 5090 just authored — result envelope routes back on the same Commands/Events bus.

Once events/command-completed (gap report priority 3) lands, build/test attribution becomes observable in real time — closing the loop from "I built this" to "the grid knows I built this." See [[alignment-via-substrate-economics]].

Test plan (29/29 pass)

parse_build_messages (5):
  parse_build_extracts_errors_with_codes_and_spans ... ok
  parse_build_separates_warnings_from_errors ... ok
  parse_build_ignores_non_diagnostic_reasons ... ok
  parse_build_tolerates_non_json_lines ... ok
  parse_build_handles_diagnostic_without_primary_span ... ok

parse_test_output (5):
  parse_test_extracts_passing_counts_from_summary ... ok
  parse_test_captures_failure_names_in_order ... ok
  parse_test_aggregates_across_multiple_test_binaries ... ok
  parse_test_dedupes_failures_across_repeated_blocks ... ok
  parse_test_empty_output_returns_zero_counts_not_error ... ok

parse_summary_counts (2):
  summary_counts_handles_filtered_out_field ... ok
  summary_counts_handles_failed_verdict ... ok  ← caught the verdict-prefix bug

timeout (2):
  timeout_uses_default_when_none_provided ... ok
  timeout_clamps_to_max_when_request_exceeds_it ... ok

types (5): round-trip, defaults, optional-omission, lib_only, failure-order

dispatch (2):
  config_advertises_cargo_prefix ... ok
  handle_command_rejects_unknown_command_loud ... ok

end-to-end (1):
  end_to_end_subprocess_pipeline_works ... ok  ← real `cargo --version`

concurrency stress (1):
  concurrent_cargo_invocations_dont_corrupt_subprocess_pipeline ... ok
  ← 8 parallel real cargo invocations, multi-thread tokio

ts-rs exports (6): all 6 binding files generated

What this PR does NOT do

• Does NOT add TS wrapper commands. Rust ServiceModule + IPC bridge is canonical per rust-is-the-core-node-is-the-shell
• Does NOT stream output line-by-line. Returns single envelope at end. Streaming = gap report priority 4 (needs Stream cell shape)
• Does NOT manage per-persona workspaces. Optional working_dir (default cwd); per-persona workspace isolation is orthogonal
• Does NOT depend on libtest's unstable JSON output. Parses stable human-readable format. Can upgrade when JSON stabilizes
• Does NOT scaffold via generate/module invocation. Hand-authored matching v2 template shape exactly. Future PR can swap in literal generator invocation

Stacks on

canary directly. Independent of any open chain.

References

• docs/planning/PERSONA-AS-DEVELOPER-GAP.md — Priority 2 (Priority 1 was feat(modules/code): code/exists + code/list + code/glob — filesystem introspection cluster #1501)
• docs/architecture/COMMAND-INFRASTRUCTURE-FIELD-MANUAL.md §3 + §4
• docs/architecture/MODULE-CATALOG.md §0 — new cargo row to add when this lands
• Memories: three-primitives-commands-events-persona, alignment-via-substrate-economics, continuum-thesis-airc-is-the-medium

Next up

Priority 3 from the gap report: events/command-completed — restores RTOS-brain doctrine by emitting completion events on the bus instead of forcing blocking polls. Largest scope of the three; touches dispatch hot path. Separate PR after this lands.