bug: --csv / --xlsx silently ignore --output

Summary

--output FILE is only consumed by the --txt branch. The --csv and --xlsx branches hardcode their filenames (f"{username}.csv" at sherlock.py:846, f"{username}.xlsx" at :927) and never consult args.output.

Observed behavior (all offline-reproducible via a stubbed sherlock())

Invocation	Result
--csv --output out.csv daniel	out.csv never created; real CSV lands at daniel.csv in CWD
--csv --txt --output out.csv daniel	out.csv contains the TXT report; real CSV still at daniel.csv
--xlsx --output out.xlsx daniel	out.xlsx never created; real xlsx at daniel.xlsx
--xlsx --txt --output out.xlsx daniel	out.xlsx is a plain-text file with txt content, not a spreadsheet
control: --txt --output out.txt	honored correctly

Notes

• Adjacent to but distinct from #2992 ("--output ignored unless --txt passed"): here --txt IS passed and the csv/xlsx output is still silently written elsewhere.
• ⚠️ Interaction warning: open PRs #3083/#3039/#2993 (all proposing the same gate change) modify the txt gate to if args.output_txt or args.output: — if merged, --csv --output out.csv would additionally write txt content into out.csv even without --txt. That widening should be considered in whatever design is chosen here.

Design questions for maintainers

With --csv --txt --output FILE:

1. Honor --output for every requested format (both writes target the same path — last-write-wins corruption)?
2. Refuse the combo (exit 1, like the existing --output/--folderoutput conflict guard at ~728-730)?
3. Derive per-format paths from the --output stem?

Happy to implement whichever is preferred — requesting direction before PRing, since the correct semantics are a product decision.

Environment: macOS, Python 3.13, master @ 3760187 (line refs verified).