Alchemical ions don't track the ligand's charge for non-linear lambda schedules #262

Alchemical ions are always given their own standard_morph schedule, so their charge changes linearly over the whole λ range. Intermediate states are then neutral only if the ligand's net charge also changes linearly, which isn't true for the ABFE (decharge stage), ring_break_morph (final morph stage), charge_scaled_morph or custom schedules. This defeats the purpose of the co-alchemical ion and adds electrostatic variance to windows that aren't spaced for it.

By default, the ion should use the ligand's schedule so that the two change their charges together. The ABFE schedules need a dedicated ion schedule: linear morph during decharge, then hold at the final values, because fix_epsilon would hold the ion's epsilon at the water value. For custom schedules, reuse the ligand schedule only if it takes the ion's charge, sigma and epsilon from their initial to their final values. Otherwise, fall back to standard_morph with a warning.