to_mmcif, to_pdb: keep link atoms present only as altlocs

[CV-GPhL]

A link partner can have no altloc - a PDB LINK record with a blank altloc
column, or an mmCIF _struct_conn row with pdbx_ptnrN_label_alt_id ? - while
the atom it names exists only in alternative conformations. find_cra() matches
altlocs strictly ('\0' matches only atoms without an altloc), so the writers
do not find the atom:

• _struct_conn: ? for the atom name, ptnr2_symmetry and pdbx_dist_value
• LINK: blank atom name, symmetry and distance
• SSBOND: the record is not written at all

Example

A zinc bound to HIS 64 NE2, which has two conformers; the LINK gives no altloc:

import gemmi

PDB = """\
LINK        ZN    ZN A 301                 NE2 HIS A  64     1555   1555  2.10
CRYST1   30.000   30.000   30.000  90.00  90.00  90.00 P 1
ATOM      1  NE2AHIS A  64      12.100  10.000  10.000  0.60 20.00           N
ATOM      2  NE2BHIS A  64      12.600  10.000  10.000  0.40 20.00           N
HETATM    3 ZN    ZN A 301      10.000  10.000  10.000  1.00 20.00          ZN
"""
st = gemmi.read_pdb_string(PDB)
print(st.make_pdb_headers().splitlines()[1])
tags = ['id', 'ptnr1_label_atom_id', 'ptnr2_label_atom_id', 'pdbx_dist_value']
print(*st.make_mmcif_document()[0].find('_struct_conn.', tags)[0])

Before (0.7.5):

LINK        ZN    ZN A 301                     HIS A  64     1555
metalc1 ZN ? ?

After:

LINK        ZN    ZN A 301                 NE2 HIS A  64     1555   1555  2.10
metalc1 ZN NE2 2.1000

The same for a disulfide read from mmCIF (the SSBOND record was lost)

CYS 5 SG has conformers A and B; the _struct_conn row has alt_id ?.

import gemmi

CIF = """\
data_example
_cell.length_a 30
_cell.length_b 30
_cell.length_c 30
_cell.angle_alpha 90
_cell.angle_beta 90
_cell.angle_gamma 90
_symmetry.space_group_name_H-M 'P 1'
loop_
_atom_site.group_PDB
_atom_site.id
_atom_site.type_symbol
_atom_site.label_atom_id
_atom_site.label_alt_id
_atom_site.label_comp_id
_atom_site.label_asym_id
_atom_site.label_seq_id
_atom_site.Cartn_x
_atom_site.Cartn_y
_atom_site.Cartn_z
_atom_site.occupancy
_atom_site.B_iso_or_equiv
_atom_site.auth_seq_id
_atom_site.auth_asym_id
ATOM 1 S SG A CYS A 5  10.00 10.00 10.00 0.70 20 5  A
ATOM 2 S SG B CYS A 5  10.00 11.60 11.60 0.30 20 5  A
ATOM 3 S SG . CYS A 20 12.04 10.00 10.00 1.00 20 20 A
loop_
_struct_conn.id
_struct_conn.conn_type_id
_struct_conn.ptnr1_label_asym_id
_struct_conn.ptnr1_label_comp_id
_struct_conn.ptnr1_label_seq_id
_struct_conn.ptnr1_label_atom_id
_struct_conn.pdbx_ptnr1_label_alt_id
_struct_conn.ptnr2_label_asym_id
_struct_conn.ptnr2_label_comp_id
_struct_conn.ptnr2_label_seq_id
_struct_conn.ptnr2_label_atom_id
_struct_conn.pdbx_ptnr2_label_alt_id
disulf1 disulf A CYS 5 SG ? A CYS 20 SG ?
"""
st = gemmi.read_structure_string(CIF)
print(st.make_pdb_headers())

Before (0.7.5): only the CRYST1 line - no SSBOND.

After:

SSBOND   1 CYS A    5    CYS A   20                          1555   1555  2.04
CRYST1   30.000   30.000   30.000  90.00  90.00  90.00 P 1

Change

New find_connection_atoms(st, con) (declared in to_pdb.hpp, next to
use_hetatm()): find_cra() for both partners, and if a partner without altloc
names an atom that is present only in alternative conformations, the closest
pair of atoms of the same conformer - as complete_ssbond() already does when
reading SSBOND records. It is used by write_struct_conn() and by the SSBOND
and LINK writers.

• No altloc is added: the alt_id / altloc column stays as the Connection has it
  (? / blank for such a partner).
• Links whose atoms are found by the strict lookup are written as before.
• Links whose atoms are really absent (LINKR gap) are written as before.

Tests

tests/test_mol.py: test_link_to_altloc_atoms and
test_ssbond_to_altloc_atoms, both failing without the change. Locally the
Python tests and cpptest pass.

Not in this PR: whether such a link should become one link per conformer, as
the wwPDB writes them (one _struct_conn row per altloc) - I will raise that
separately in an issue.