vecann: fix concurrent download race in dataset cache

[mw5h]

Summary

• Fixed a race condition causing 100% nightly failure of vecindex/dbpedia-100k/nodes=3/prefix=0 on master.
  The prefix=0 and prefix=3 test variants run concurrently and both download
  the same dataset to the same cache directory using a fixed temp file path,
  causing concurrent writers to corrupt each other's data.
• Changed downloadAndUnzip to use os.CreateTemp for unique temp files per
  downloader, with atomic os.Rename to install the extracted file at the
  destination. The cached file is now either absent or complete — never truncated.
• Removed ResetCache:true from the vecindex roachtest (added in 2794f07 as
  a workaround for corrupt cache files). With atomic extraction the workaround is
  no longer needed, and removing it eliminates the redundant 443MB download that
  was triggering the race.

Fixes: #163471
Fixes: #159333

Test plan

• Ran vecindex/random-s/nodes=1/prefix=0 locally with --local — passed (2250s).

[trunk-io]

😎 Merged successfully - details.

[cockroach-teamcity]

This change is 

[mw5h]

@mw5h made 1 comment.
Reviewable status: complete! 0 of 0 LGTMs obtained.

---

pkg/workload/vecann/datasets.go line 349 at r1 (raw file):

	// Atomic rename: the destination is either absent or contains the complete
	// extracted file. Concurrent processes performing the same download will
	// each rename their own temp file, with the last writer winning.

We already said this above.

[yuzefovich]

Nice find!

@yuzefovich reviewed 2 files and all commit messages, and made 3 comments.
Reviewable status: complete! 1 of 0 LGTMs obtained (waiting on mw5h, nameisbhaskar, and williamchoe3).

---

pkg/cmd/roachtest/tests/vecindex.go line 256 at r2 (raw file):

	loader := vecann.DatasetLoader{
		DatasetName: opts.dataset,
		ResetCache:  true,

nit: we can now remove ResetCache option since it's no longer used anywhere.

---

pkg/workload/vecann/datasets.go line 303 at r2 (raw file):

	writer.OnProgress = dl.OnDownloadProgress
	_, copyErr := io.Copy(&writer, reader)
	_ = reader.Close()

nit: why are we ignoring this error but not on tempZip.Close? A quick comment would be helpful.