Fix mergeIntrascale dropping (1,1) pairs in ring-breaking perturbations

[lohedges]

Fixes a bug in the new SireIO::mergeIntrascale function where the helper lambda copyIntrascale skipped writing CLJScaleFactor(1, 1) pairs because of an explicit guard against the default value. In the two-pass algorithm used to build each merged end-state intrascale, the first (ghost-topology) pass could write non-(1, 1) values for shared-atom pairs, and the second (correct end-state) pass would then fail to overwrite them when the correct value was (1,1). This caused ring-breaking perturbations (e.g. cyclopentane → cyclohexane) to produce too few changed OpenMM exceptions; 6 instead of 18, leading to missing non-bonded interactions and simulation instabilities. The fix adds a copy_all flag to copyIntrascale and passes it as true on the second call for each end state, ensuring the correct end-state values always win regardless of whether they are (1, 1) or not.

No CHANGELOG entry is needed since this new function is not (yet) part of a release. The update has been tested locally and a regression test will be provided via BioSimSpace in a companion PR.

• I confirm that I have merged the latest version of devel into this branch before issuing this pull request (e.g. by running git pull origin devel): [y]
• I confirm that I have permission to release this code under the GPL3 license: [y]