kenref: support more than one rank per replica (#1443)

[aalhossary]

Description

Follow-up to #1443 for KENREF, in the same way as #1486 for TTMETAD. KENREF refused to run when a replica spans several ranks; this makes it work.

calculate() read the replica count and index from multi_sim_comm on every step and on every rank. A rank that is not the first of its replica only has the MPI_COMM_SELF stand-in, which reports "replica 0 of 1", so it would have applied a one-replica restraint to the atoms it owns without any error. That is why the module refused this case until now.

• The replica count and index come from getNumberOfReplicas() / getReplicaIndex() (Replica count and index on every rank of a replica (#1443, part C) #1481), once, at setup.
• Only the first rank of each replica gathers and scatters over multi_sim_comm. The derivatives are then broadcast over comm. The broadcast is required: every rank applies forces to the atoms it owns, so a rank without derivatives would apply none.
• The energy is printed once per job, not once per rank.

Merge order. This needs the matching KEnRef change, Smith-Group/KEnRef@b4aa301 on release/2.x. The new code uses members that only that KEnRef declares, and a PLUMED and a KEnRef that do not match fail to compile; they never build and give wrong forces. This PR deliberately pins .ci/install.kenref to release/2.x, which contains that commit, so CI builds the KEnRef it needs. Between approval and merge (or in a follow-up), I will move v2.0.0 to that commit and return the pin to v2.0.0.

Target release

I would like my code to appear in release master (v2.11)

Type of contribution

• changes to code or doc authored by PLUMED developers, or additions of code in the core or within the default modules
• changes to a module not authored by you
• new module contribution or edit of a module authored by you

Copyright

• the module I added or modified contains a COPYRIGHT file with the correct license information. Code should be released under an open source license. I also used the command cd src && ./header.sh mymodulename in order to make sure the headers of the module are correct.

Tests

• I added a new regtest or modified an existing regtest to validate my changes.
• I verified that all regtests are passed successfully on GitHub Actions.

New kenref/rt-kenref-sigma-multi-dd: 2 replicas × 2 ranks with driver --debug-dd yes, using real two-member data. The replicas start from different frames, so their forces differ (by up to 4.56). The COLVAR files and all four force dumps are byte-identical to the same run with one rank per replica. Removing the broadcast makes all four force files fail. rt-kenref-sigma is unchanged and passes.

Also checked locally:

• With --enable-debug (Raise an error in debug builds when multi_sim_comm is used where it is a stand-in (#1443) #1484), the new test never uses the stand-in. Re-adding an ungated gather trips the Raise an error in debug builds when multi_sim_comm is used where it is a stand-in (#1443) #1484 error on exactly the two non-first ranks. The -debug- CI jobs do not build kenref, so this check cannot run on this module in CI.
• In a 500-step GROMACS run with 2 replicas × 2 ranks, the energy at step 0 is identical to the run with one rank per replica. At step 10 the two differ by 2e-6 relative, which is as much as two repeats of the DD run differ from each other. Without the broadcast, the difference at step 10 is 3.5e-2 and the mean energy is 44% higher.