Keep the 0.9.39 cdxt command lines working (deprecated)

[malteos]

Keeps the command lines of cdx_toolkit 0.9.39 working on top of #79. Without this PR, every documented old command line fails with "unrecognized arguments", for example cdxt --cc --limit 10 iter 'commoncrawl.org/*', --crawl 3 … iter, size and warc. This includes the examples in Common Crawl's own whirlwind docs.

What changes

cdxt recognizes an old command line in either of two cases:

• the command is iter, size or warc;
• options other than -h, -V or -v come before the command.

Such command lines are parsed with the 0.9.39 grammar, so mistakes still exit with status 2, and they run with the new code. stderr gets one line with a suggested new command:

$ cdxt --cc --limit 10 iter 'commoncrawl.org/*'
cdxt: this command line uses the old syntax, which is deprecated and will be removed in a future version; suggested replacement: cdxt index cc --url 'commoncrawl.org/*' --limit 10 --output-format csv --fields status,timestamp,url (see …)
status 200, timestamp 20250804103329, url https://www.commoncrawl.org/

• iter:
  • The old text, --jsonl and --csv output and the default fields url,status,timestamp are unchanged.
  • Captures come from the CDX API backend, which sends the same requests as before.
  • --get uses CDXFetcher.get() (default limit 1000).
  • --closest without --get uses paged requests, as before.
• size: runs CDXFetcher.get_size_estimate(). --details works, and the result is not capped by --limit.
• warc:
  • The captures stream into repackage() with sort=False, so memory use stays flat and index paging overlaps with fetching.
  • --url-fgrep / --url-fgrepv apply after the limit.
  • --ia replays from the Internet Archive without --wb.
• The new-arguments cdxt iter alias is removed (it was never released), and so is cli/deprecated.py.

Accepted differences from 0.9.39

• warc:
  • writes TEST-001.warc.gz instead of TEST-000000.extracted.warc.gz;
  • copies records byte-for-byte, without the WARC-Source-* headers;
  • writes no file when nothing is selected;
  • exits with status 1 if a capture could not be fetched.
• iter --all-fields --csv is now a usage error. In 0.9.39 it failed with a traceback.

Tests

• tests/test_cli_legacy.py: the 0.9.39 command lines, run against the responses that were recorded for them in tests/data/mock_responses/test_cli/. They check the output format, --get, size, size --details, and the WARC files and warcinfo for --cc, --ia, the prefix options and an empty selection.
• tests/unit/test_cli_legacy.py covers:
  • detection of old command lines;
  • strict errors, with no requests made;
  • source precedence;
  • the formatters;
  • the fetcher paths;
  • the warc content source, sort=False and exit status 1;
  • that every suggested replacement parses with the new parser.