Fix pandas 2.x ValueError in monomer SDF loading

[anagnorisis2peripeteia]

Summary

Fixes #18

Pandas 2.x introduced Arrow-backed storage for string columns. When
get_monomer_info() and _load_monomer_sdf() attempt to write a parsed
Python list into these columns via df.loc[idx, col] = [...], pandas 2.x
raises:

ValueError: Must have equal len keys and value when setting with an iterable

because it interprets the list as multiple row values rather than a single
list object for one cell.

Fix

Two changes, applied in both sequence.py and monomerlib.py:

1. Cast the three list-valued columns (m_Rgroups, m_RgroupIdx,
   m_attachmentPointIdx) to object dtype before writing into them.
   Only these three columns are cast — the rest of the DataFrame retains
   its pandas 2.x type optimisations.

2. Use df.at[idx, col] (single-cell assignment) instead of
   df.loc[idx, col] (which pandas 2.x misinterprets as a broadcast
   when given an iterable).

Note on duplication

get_monomer_info() in sequence.py and _load_monomer_sdf() in
monomerlib.py implement the same SDF loading and list-parsing logic
independently. This fix is applied to both. A follow-up refactor could
consolidate them into a single shared loader to avoid this kind of
divergence in future — happy to open a separate PR for that if useful.