Fix PmagPy_MagIC.ipynb under pandas 3.0 and current PmagPy/MagIC API - #866

The oersted parameter introduced in 1deb048 (2017) was never branched
on, making the /10 Gauss-to-Tesla correction unconditional; the parameter
itself was later removed in 533e8dc but the unconditional correction
remained. Restore oersted=True as the default (matching the CIT format
convention where AF step values are recorded in Gauss) and gate the
conversion on the flag so callers with mT-formatted .sam files can pass
oersted=False to get correct Tesla values. Replace the redundant -oe CLI
flag with -mT.

…RE converters

pandas 3.0's strict str dtype rejects .loc[mask, col] = <float> partial
assignments into string-initialized columns. In iodp_jr6_lore, treat_temp
and treat_dc_field are initialized as strings and then partially overwritten
with float values for TD and ARM rows; in iodp_dscr_lore the same pattern
fires on treat_dc_field for ARM rows. Cast these columns to object dtype
immediately after their string init so the later partial assigns succeed.

save_plots had dir_path=None as default, then unconditionally did
os.path.join(dir_path, fname), which raises TypeError when no caller
forwards dir_path. Most callers in ipmag (aniso_magic, zeq_magic plot
saves, etc.) omit the argument, so they hit this whenever not running
on the MagIC server and no dpi kwarg is passed. Change the default to
"." so the docstring's promise ("directory in which to save plots") is
the actual fallback behavior.

Restore sites.txt and magic_contribution_16533.txt, which were deleted as
collateral damage in commit 44293e8 (2021-06-20) when an in-progress
extract_results.ipynb was added; the companion notebook was itself deleted
13 minutes later in 00f9151 but these data files were never put back.
The PmagPy_MagIC.ipynb download_magic section has been quietly broken on
that missing sites.txt ever since. sites.txt is restored from 9329ca3
(its last modification, 2020-09-04) and magic_contribution_16533.txt from
2ef7d21~1 (the last commit before its 2020-09-04 deletion).

…ples

- Cell 24 (convert_ages): cast frame to object before fillna('') so
  pandas 3.0's strict str dtype accepts empty strings in numeric columns
- Cell 57 (private download example): replace stale credentials
  (contribution 16960 / fa06255d-ec45-473f-9f33-fa2a9bb6d8d8) with a
  currently-shared contribution (20672) since the original was no longer
  accessible via the MagIC API
- Cell 60 (upload_magic): forward input_dir_path explicitly (the function
  was split into separate input/output paths) and pass validate=False
- Cell 78 (markdown): note that the next cells are skipped in the rendered
  documentation version of the notebook
- Cell 231 (cleanup): early-return in remove_magic_files when the directory
  is missing, wrap remaining os.remove calls in try/except FileNotFoundError,
  add 2g_asc_magic and 2g_bin_magic parents to the directory list, sweep
  leftover magic_contribution_*.txt at the project root, and clean
  data_files/download_magic/ while preserving the two example files

The new test_convert_cit.py covers basic conversion of bl9-1.sam to the
five MagIC 3.0 tables (record counts, headers, sitename/locname
propagation, specimen↔measurement consistency), the oersted flag's Gauss
vs milliTesla interpretation of AF demag steps, and the actionable-error
return tuple for invalid inputs. Mirrors and extends the unittest-style
coverage in pmagpy_tests/test_imports3.py, which is the older suite that
CI does not run.

Several pmagpy/test/test_ipmag_*.py files import pmagpy.ipmag, which
conditionally pulls in cartopy via pmag.import_cartopy(). When cartopy
isn't installed in the CI image, pytest still collects the module but
downstream cartopy-using paths break test collection during import.
Adding cartopy to the workflow's pip install line resolves the missing
dependency so the suite collects cleanly.

pmagpy/convert_2_magic.py imports pytz at module top, so the new
test_convert_cit.py fails collection on a CI image where pytz isn't
already pulled in transitively. Adds pytz alongside cartopy in the
workflow's pip install line.