stream: cut promise churn in webstreams hot paths

[mcollina]

Twelfth round of pure-JS webstreams optimizations, following #64890.

Profiling pipeTo at the default highWaterMark showed roughly 10% of the profile in queueMicrotask plus its native binding (one call per pump batch), a fresh writer.ready promise record plus reaction per backpressure flip on the pipeThrough shape, and an implicit async-wrapper promise per sink.write/source.pull invocation. Three changes, one commit:

• Non-thenable algorithm fast path. User sink.write and source.pull callbacks are wrapped without coercing their result into a promise. A non-thenable result (the common synchronous case) means fulfillment is guaranteed and no then() lookup is observable, so the fulfilled reaction is enqueued through a single shared resolved promise at the exact microtask position the coerced promise's reaction would have had. Thenable results go through PromiseResolve(), matching the reference implementation's promiseCall (identity for native promises).
• pipeTo parks on backpressure via a ready-record hook. The pump installs a record that duck-types the writer's lazily-materialized [[readyPromise]] record; writableStreamUpdateBackpressure resolving it re-enters the pump directly. This removes the per-flip promise record + reaction and the per-batch queueMicrotask. The backpressure state field is now published before the ready record is resolved so the hook observes the new value (the resolve of a real ready record only settles a promise, so the reorder is unobservable otherwise).
• Chunk-forwarding microtasks via a shared resolved promise. The pipeTo/tee forwardChunk hops and the pump's between-batch yield use a reaction on the shared resolved promise instead of queueMicrotask, which enqueues at the same position with less overhead.

A separate first commit fixes the pipe-to.js benchmark: the highWaterMark values were passed inside the underlying source/sink dictionaries where they are ignored, so all 16 configurations measured the identical workload at the default highWaterMark of 1. The strategies are now passed as the constructors' second argument, with the matrix covering the default (1) and buffered (1024, 4096) configurations.

Benchmark results with the fixed benchmark (30 runs):

                                                                       confidence improvement accuracy (*)    (**)   (***)
webstreams/pipe-to.js highWaterMarkW=1 highWaterMarkR=1 n=500000                *      9.34 %       ±8.20% ±10.90% ±14.20%
webstreams/pipe-to.js highWaterMarkW=1 highWaterMarkR=1024 n=500000            **     12.81 %       ±8.69% ±11.57% ±15.06%
webstreams/pipe-to.js highWaterMarkW=1 highWaterMarkR=4096 n=500000            **     14.09 %       ±8.45% ±11.24% ±14.64%
webstreams/pipe-to.js highWaterMarkW=1024 highWaterMarkR=1 n=500000             *      8.87 %       ±7.41%  ±9.86% ±12.83%
webstreams/pipe-to.js highWaterMarkW=1024 highWaterMarkR=1024 n=500000          *      9.51 %       ±8.64% ±11.50% ±14.97%
webstreams/pipe-to.js highWaterMarkW=1024 highWaterMarkR=4096 n=500000          *      8.01 %       ±7.93% ±10.55% ±13.74%
webstreams/pipe-to.js highWaterMarkW=4096 highWaterMarkR=1 n=500000             *      9.36 %       ±8.27% ±11.01% ±14.33%
webstreams/pipe-to.js highWaterMarkW=4096 highWaterMarkR=1024 n=500000         **     11.51 %       ±8.56% ±11.39% ±14.82%
webstreams/pipe-to.js highWaterMarkW=4096 highWaterMarkR=4096 n=500000          *     10.78 %       ±8.37% ±11.14% ±14.50%

The full-suite run showed no regressions in any other family; readable-read normal, tee normal, and pipeThrough passthrough spot runs also improve (~+6-18%). Verified with the WPT streams/compression/encoding suites, the full parallel webstream/whatwg test set, and a shutdown-ordering stress (abort mid-write, close with pending writes, sync-throwing and rejecting sinks, error propagation) whose event log is byte-identical to main.

[nodejs-github-bot]

Review requested:

• @nodejs/performance

[jasnell]

Couple of nits, otherwise LGTM

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.30%. Comparing base (347e266) to head (e30f373).
⚠️ Report is 8 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff            @@
##             main   #65138    +/-   ##
========================================
  Coverage   90.30%   90.30%            
========================================
  Files         759      759            
  Lines      248294   248415   +121     
  Branches    46860    46871    +11     
========================================
+ Hits       224220   224334   +114     
+ Misses      15516    15503    -13     
- Partials     8558     8578    +20     

Files with missing lines	Coverage Δ
lib/internal/webstreams/readablestream.js	98.18% <100.00%> (+0.02%)	⬆️
lib/internal/webstreams/util.js	97.89% <100.00%> (+0.17%)	⬆️
lib/internal/webstreams/writablestream.js	99.52% <100.00%> (+<0.01%)	⬆️

... and 37 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.