feat: update to MoonBit v0.10.4 / async 0.20.2 and lift the 1000 req/s rate cap

[paveg]

Summary

Update the MoonBit toolchain to v0.10.4 and all dependencies to their latest versions,
fix a pre-existing hang in the rate limiter,
and lift its 1000 req/s cap.

The toolchain and moonbitlang/async are tightly coupled and cannot be updated independently:
async 0.20.x requires the moonc v0.10.4 syntax and the new manifest format,
while moonc v0.10.4 rejects the legacy syntax used inside async 0.17.1.
That is why the dependency updates land as a single PR.

Changes

Dependency updates (mechanical)

• moonbitlang/async 0.17.1 → 0.20.2 / mizchi/tui 0.9.1 → 0.10.0 / moonbitlang/x 0.4.40 → 0.4.46
• Migrate moon.mod.json to moon.mod (DSL) — auto-migrated by moon fmt; the json form is deprecated under the new toolchain
• Switch release-please to the generic updater with x-release-please-start-version block annotations in moon.mod
• Update expect tests for the new core Show format (String tuple elements are no longer quoted)
• Migrate deprecated APIs: verify~ → trust~ (@tls.TrustedRoot), and Client::new / Server::new → struct-constructor call style.
  --insecure semantics are unchanged (insecure → NoVerification, default → SystemRoot)

Rate limiter (core)

• fix: resolve a pre-existing hang whenever --rate was set (reproducible in v0.1.9).
  The token filler loops forever, and spawn_bg waits for child tasks by default,
  so with_task_group never returned once the test finished.
  The filler is now spawned with no_wait=true, making it an auxiliary task the group cancels on termination
• feat: support rates above 1000 req/s via batch token emission.
  @async.sleep has millisecond resolution, so the previous one-token-per-sleep filler topped out at ~1000 tokens/s.
  The filler now computes how many tokens are due at the current instant against the absolute schedule and emits them as a batch,
  keeping exact long-run pacing for any rate

Verification

• moon test --target native 256/256, moon test (wasm-gc) 187/187
• End-to-end against the built-in server: -d 3s -r 3000 measured 3000.13 req/s (9,006 requests); -r 100 measured 100.40 req/s
• Confirmed the -n+-r and -d+-r hangs are gone by comparing the new binary against the v0.1.9 release binary
• TUI rendering and terminal restore verified under a pty with tui 0.10.0
• Reviewed by a separate spec-review agent (0 critical findings; the single consistency finding was addressed in b689970)

Notes

A pre-existing display bug was found during this work and is intentionally out of scope for this PR:
format_us_as_ms drops the leading zero of the fractional part, rendering 80µs as "0.8ms" instead of "0.08ms" (src/lib/reporter/format.mbt).
A follow-up PR will address it.

🤖 Generated with Claude Code

[changeset-bot]

⚠️ No Changeset found

Latest commit: b689970

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR