[Node.js] disable source maps in node heap

[IlyasShabi]

The node_heap test started failing after dd-trace-js changed profiler startup to happen earlier, during tracer initialization.

References:

• Artifact: https://github.com/DataDog/prof-correctness/actions/runs/23798552320/artifacts/6201603273
• dd-trace-js PR: feat(profiling): Start profilers synchronously within tracer initialization dd-trace-js#5906
• pprof-nodejs PR: Rewrite SourceMapper to scan directories with sourceMappingURL support pprof-nodejs#292

1. The failing runs show this error: TypeError: mapper.loadDirectory is not a function and the issue turned out to be a version mismatch. The package dd-trace is installed from master branch, but @datadog/pprof was overridden to "dev" which is an npm dist-tag, not a git branch, and it was resolving to an older pre-release build 6.0.0-pre-fdd7f20 rather than the current main branch.
   The pprof pipeline to publish dev was fixed use trusted publishing for dev releases pprof-nodejs#310 and use release publisher file to release dev tag too pprof-nodejs#311
2. Update measurements

[r1viollet]

why is this changing to 50 50 ?

[IlyasShabi]

These are the observed values and reflect exactly what the code is doing. When back to 60 I had from CI

Assertion failed: stack 'processTimers;listOnTimeout;work;b;slice' (labels=[]) should have been 66% +/- 5% of the profile but was 50% with 16% error

Assertion failed: stack 'processTimers;listOnTimeout;work;a;slice' (labels=[]) should have been 33% +/- 5% of the profile but was 49% with 16% error

[r1viollet]

I understand, but are these values expected ?
Is the scenario meant to have 1/3 2/3 split or a 50/50 split ?

[r1viollet]

  function work() {
    count += 1
    a(allocSize, refs);
    b(allocSize * 2, refs);

[IlyasShabi]

IMO the 50/50 is correct because the objects profile counts allocations, not bytes. In work() a and b are called once on every iteration.

The old 66/33 value for objects was not the expectation of the scenario. It was an observed bias from earlier runs https://github.com/DataDog/prof-correctness/blob/main/scenarios/node_heap/README.md?plain=1#L23

[r1viollet]

ah OK, this is number of objects, apologies

[r1viollet]

NIT: could requires a comment to explain that we can remove this at a specific release

[IlyasShabi]

I just checked that the default value is false not needed anymore

[r1viollet]

LGTM