Fix hScanValuesIterator to yield values instead of field names

[winklemad]

hScanValuesIterator() was a copy of hScanNoValuesIterator(): it called hScanNoValues() (which issues HSCAN ... NOVALUES, a reply that by protocol contains only field names) and yielded reply.fields. As a result it returned hash field names — byte-for-byte identical to hScanNoValuesIterator — instead of the values it is named for.

hScanIterator (entries) and hScanNoValuesIterator (field names) were both correct; only the values-only variant was wrong.

Fix

Iterate the full HSCAN, exactly like hScanIterator already does, and yield the values:

-      const reply = await this.hScanNoValues(key, cursor, options);
+      const reply = await this.hScan(key, cursor, options);
       cursor = reply.cursor;
-      yield reply.fields;
+      yield reply.entries.map(entry => entry.value);

Behavior for a hash { f1: 'v1', f2: 'v2' }

	before	after
hScanValuesIterator yields	['f1', 'f2'] (field names)	['v1', 'v2'] (values)

Fixes #3343

---

Note

Low Risk
Small, localized bugfix to a misnamed iterator with a new integration test; behavior change only affects callers that relied on the incorrect field-name output.

Overview
hScanValuesIterator was implemented like hScanNoValuesIterator: it used hScanNoValues (HSCAN … NOVALUES) and yielded reply.fields, so callers got field names instead of values.

The iterator now uses full hScan (same pattern as hScanIterator) and yields reply.entries.map(entry => entry.value) per page.

An integration test hScanValuesIterator asserts a 100-field hash returns all expected valueN strings.

^{Reviewed by Cursor Bugbot for commit b57c4f5. Bugbot is set up for automated code reviews on this repo. Configure here.}

[nkaradzhov]

@winklemad thanks for the PR!
One change needed before merge: please add a regression test in packages/client/lib/client/index.spec.ts, following the existing hScanIterator / hScanNoValuesIterator tests (testUtils.testWithClient), asserting that the iterator yields the hash values. Happy to merge once that's in.

[winklemad]

Thanks @nkaradzhov — added the regression test in b57c4f5, mirroring the existing hScanNoValuesIterator test. It uses distinct field/value pairs (value${i}) and asserts the iterator yields the hash values, so it fails against the old field-name behavior and passes with the fix. Re-requesting review whenever you have a moment.

[nkaradzhov]

Thanks @winklemad, looks good!