fix: handle PAX and GNU metadata entry types in TAR extraction

[bug-ops]

Summary

The catch-all _ arm in to_entry_type() incorrectly mapped PAX headers and GNU metadata to SecurityViolation, breaking extraction of PAX archives.

• Skip XHeader, XGlobalHeader, GNULongName, GNULongLink as format-internal metadata
• Map Continuous and GNUSparse to File (real data entries)
• Apply same filtering in list.rs for extract()/list_archive() consistency

Test plan

• Tests for all new code paths (PAX skip, GNU metadata skip, sparse, continuous, unknown byte rejection)
• 635 tests pass, fmt/clippy clean

Closes #69

[bug-ops]

Verified with the reproducer from the issue description (built from #70):

  import io, tarfile, tempfile, exarch

  buf = io.BytesIO()
  with tarfile.open(fileobj=buf, mode="w:gz") as tar:
      pax = tarfile.TarInfo(name="././@PaxHeader")
      pax.type = b"g"
      pax_data = b"16 comment=hi\n"
      pax.size = len(pax_data)
      tar.addfile(pax, io.BytesIO(pax_data))

      info = tarfile.TarInfo(name="hello.txt")
      content = b"hi"
      info.size = len(content)
      tar.addfile(info, io.BytesIO(content))

  with tempfile.TemporaryDirectory() as tmp:
      archive_path = f"{tmp}/test.tar.gz"
      with open(archive_path, "wb") as f:
          f.write(buf.getvalue())
      report = exarch.extract_archive(archive_path, tmp)
      print(report)

ExtractionReport(files=1, dirs=0, symlinks=0, bytes=2, duration=0ms, skipped=0, warnings=0)