fix(pl-middle-layer): stop the render test racing a cold block fetch

[DenKoren]

Stacked on #1836 (chore/build-on-windows) — review that one first; this branch
only adds the commits on top. Base retargets to main once #1836 merges.

Fixes the intermittent test JS render enter numbers failure seen on #1836,
which is a pre-existing flake on main, not something that branch introduced.

Decisions

The bug is the scope of the abort signal, not its value. awaitBlockDone
builds one AbortSignal.timeout(timeout) outside the while loop and reuses
it for every overview.awaitChange(...). So the value is the budget for the
entire wait, not for one change notification — a name that reads like a per-wait
timeout but behaves like a deadline. Both copies of the helper have this shape.

The old budgets could not cover the work they were waiting on. The live
caller in render.test.ts adds a block by from-registry-v1 from
https://block.registry.platforma.bio/releases and then runs it, so its 2000ms
had to cover a cold remote fetch plus execution on a CI runner. That is a race
against the network, which is why it failed intermittently and why the failure
looked like nothing in particular.

Both defaults are now 10s. render.test.ts was 2000ms and
DEFAULT_AWAIT_BLOCK_DONE_TIMEOUT in sdk/test was 5000ms. 10s sits well
inside the 80s testTimeout of the middle-layer suite, so a genuinely stuck
block still fails the test rather than hanging it, and callers that need longer
keep passing their own value (15s–50s across the repo).

Raised the defaults rather than restructuring the helper. A progress
watchdog — reset the deadline on each observed change, with an overall cap —
buys nothing here: during the fetch there is no overview change to reset on,
which is exactly the interval that has to be tolerated, so the idle bound would
have to be just as generous as the total. It would also be a new pattern in a
repo that consistently uses a single overall budget.

The abort message now says what was pending. It was a bare "Aborted.",
which is what made the CI failure take a log dig to identify. It now names the
block, its last calculationStatus and the budget, and dumps blockOverview
beside the block state — matching what the sdk/test copy already prints.

Key files to review

1. lib/node/pl-middle-layer/src/middle_layer/render.test.ts — the signal's
   scope is the whole point, so read the comment against the loop below it.
2. sdk/test/src/test-block.ts — same shape, published package; the default
   here is exposed to every block test that does not pass its own timeout.
3. .changeset/await-block-done-default-timeout.md — patch bump for
   @platforma-sdk/test.

Verification

types:check, formatter:check and linter:check pass on both packages. I
have not reproduced the flake locally — it needs a live backend and a cold
registry fetch, and a green local run would not be evidence either way for a
timing race. The claim rests on the read of the code: a 2s deadline spanning a
network download, with every sibling caller at 15s or more.

[changeset-bot]

🦋 Changeset detected

Latest commit: 6060a2c

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 1 package

Name	Type
@platforma-sdk/test	Patch

Not sure what this means? Click here to learn what changesets are.

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

[greptile-apps]

Misleading caller description

The comment says other callers of “this helper” pass 15–50-second timeouts, but this helper’s only active caller uses the default. The similarly named helpers elsewhere are separate implementations. Describing them as analogous helpers would make the timeout rationale accurate and easier to verify.

Suggested change

	* download on a CI runner. Other callers of this helper across the repo pass
	* 15-50s for the same reason.
	* download on a CI runner. Analogous helpers elsewhere in the repo use
	* 15-50s for the same reason.

Prompt To Fix With AI

This is a comment left during a code review.
Path: lib/node/pl-middle-layer/src/middle_layer/render.test.ts
Line: 40-41

Comment:
**Misleading caller description**

The comment says other callers of “this helper” pass 15–50-second timeouts, but this helper’s only active caller uses the default. The similarly named helpers elsewhere are separate implementations. Describing them as analogous helpers would make the timeout rationale accurate and easier to verify.

```suggestion
 * download on a CI runner. Analogous helpers elsewhere in the repo use
 * 15-50s for the same reason.
```

---

For each issue above, determine whether it is valid and should be fixed. If so, fix it directly.

Note: If this suggestion doesn't match your team's coding style, reply to this and let me know. I'll remember it for next time!