Fix dead wikitext-103 S3 link in quicktour.mdx

[RudrenduPaul]

Closes #1625

What this PR does

Fixes the broken wget URL in the "Build a tokenizer from scratch"
section of docs/source-doc-builder/quicktour.mdx. The link pointed at:

https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-103-raw-v1.zip

which has been dead since the research.metamind.io bucket was taken
down (confirmed still broken today — see Verification below). Anyone
following the quicktour hits a dead link on the very first command.

Change

docs/source-doc-builder/quicktour.mdx: replaced the dead S3 URL with
the canonical wikitext-103-raw-v1.zip mirror hosted on the Hugging
Face Hub (huggingface.co/datasets/mattdangerw/wikitext-103-raw). The
archive contents and internal layout are identical, so the existing
unzip wikitext-103-raw-v1.zip step and downstream paths in the rest of
the tutorial are unaffected.

# before
wget https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-103-raw-v1.zip

# after
wget https://huggingface.co/datasets/mattdangerw/wikitext-103-raw/resolve/main/wikitext-103-raw-v1.zip

Background

• PR Update quicktour.mdx re: Issue #1625 #1846 (merged) already fixed the related Salesforce blog link and
  an intra-doc link elsewhere in this same file.
• The wget URL itself was left broken and is tracked by both Tokenizer Quickstart Tutorial: Broken Links #1625 and
  the duplicate wikitext-103-raw-v1.zip is not available on the amazonaws anymore #1683.
• A prior PR (Replace dead wikitext s3 link in quicktour with HF dataset mirror #2064) proposed the same HF Hub mirror fix and was never
  merged — its author self-closed it without a stated reason (no
  maintainer review, no CI run, no comments) on 2026-06-06. Both Tokenizer Quickstart Tutorial: Broken Links #1625
  and wikitext-103-raw-v1.zip is not available on the amazonaws anymore #1683 are still open on main today, and no other PR currently
  addresses them, so this re-proposes the same fix.

Verification

Confirmed the old link is still dead and the replacement resolves and
serves the real archive:

$ curl -sI https://s3.amazonaws.com/research.metamind.io/wikitext/wikitext-103-raw-v1.zip
HTTP/1.1 301 Moved Permanently
(bucket redirect does not resolve to usable content)

$ curl -sI -L https://huggingface.co/datasets/mattdangerw/wikitext-103-raw/resolve/main/wikitext-103-raw-v1.zip
HTTP/2 302
...
HTTP/2 200
content-type: application/zip
content-length: 191984949
content-disposition: inline; filename*=UTF-8''wikitext-103-raw-v1.zip; filename="wikitext-103-raw-v1.zip"

The Hub redirects to its CDN and serves the full ~192 MB zip with the
expected filename and content type.

Scope

Docs-only, single file changed. No code, tests, or build changes.

Note: Claude Code was used to assist in drafting this change. All
changes were reviewed by the submitter.