fix: robust 429 retry in get_data.sh prevents 0-byte test data files

[Copilot]

massiveproxy.gnps2.org rate-limits to ~10 req/min. The previous wget-based download loop (3 retries, 5s waitretry) exhausted all attempts within the same rate-limit window and — critically — wget --output-document always creates the output file, even on total failure. Tests then received 0-byte files and died with lxml.etree.XMLSyntaxError: no element found rather than a meaningful download error.

Changes

• Switch wget → curl with HTTP status capture (-w "%{http_code}"), enabling reliable 429 detection at the shell level
• 429-aware backoff: waits 60s on first rate-limit hit, adding 30s per subsequent retry (up to 8 attempts)
• No stale empty files: failed attempts rm -f the output before sleeping, eliminating the 0-byte artifact that silently poisoned tests
• Non-empty validation: success requires both HTTP 200 and -s (non-empty file) check
• 7s inter-download sleep maintained to stay within the ~10 req/min budget on happy-path runs

download() {
    ...
    http_code=$(curl -L -o "$filename" -w "%{http_code}" --silent --show-error "$url")
    if [ "$curl_exit" -eq 0 ] && [ "$http_code" = "200" ] && [ -s "$filename" ]; then
        sleep 7; return 0
    fi
    rm -f "$filename"
    if [ "$http_code" = "429" ]; then
        wait_time=$((60 + retry * 30))
        sleep "$wait_time"
    fi
    ...
}