fix: handle urllib3 LocationParseError (crash on usernames ending in .)

[SkxOverKill]

Summary

Fixes #2970 — Sherlock crashes (unhandled urllib3.exceptions.LocationParseError) when a username produces an unresolvable URL host, e.g. alice. builds https://alice..empretienda.com.ar for Empretienda AR.

requests does not wrap urllib3 LocationParseError, so it escaped the per-site error handling in get_response() and killed the entire run via a request_future.result() exception. This converts it into a per-site "Invalid URL" error (QueryStatus.UNKNOWN) so the rest of the usernames/sites continue.

Reproduction (before)

python sherlock alice. ended with:

urllib3.exceptions.LocationParseError: Failed to parse: 'alice..empretienda.com.ar', label empty or too long

After

alice. is recorded against each site as before; the Empretienda AR site is reported as an error instead of crashing the whole search.

Changes

• sherlock_project/sherlock.py: handle urllib3.exceptions.LocationParseError in get_response() alongside the other per-site errors.
• pyproject.toml: declare the now-direct urllib3 dependency.
• tests/test_probes.py: offline regression test test_location_parse_error_does_not_crash asserting the exception is surfaced as an error, not re-raised.

Verification

• pytest -m "not online" offline suite passes (including the new regression test).
• ruff check introduces no new violations vs. HEAD baseline.