fix(structure): reject atom-to-self bonds in BondList and chem_comp_bond writer

[haoyu-haoyu]

Fixes #812.

A `BondList` should encode chemistry, so a bond whose two atom indices refer to the same atom is meaningless. Today both the constructor and `add_bond()` accept self-bonds silently.

Changes

• `BondList.add_bond` raises `ValueError` when the resolved positive indices are equal, including the case where one index is given positive and the other negative (e.g. `4, -1` with `atom_count=5` — both resolve to atom `4`).
• `BondList.init` performs the same check on the bulk-add path after the per-row sort, so building a `BondList` from a numpy array that contains a self-bond row raises instead of silently storing it.
• The companion concern from Bond can be created of an atom to itself #812 — that `pdbx.set_structure()` could emit a self-bond into `chem_comp_bond` when two distinct atoms in an `AtomArray` share the same `(res_name, atom_name)` annotations — is also addressed: `_set_intra_residue_bonds()` now filters bond rows where the two `atom_name` annotations are equal before deduplication, and returns `None` if no rows remain.

Tests

• `test_invalid_creation` extended with two self-bond cases (positive index pair, mixed positive/negative).
• New `test_add_self_bond_rejected` covers `add_bond()` for both index forms and verifies the list is unchanged after each rejected add.