General issues among all versions

[nadnaps]

• Halo updates of all velocities before mlsForce
• Syncing collision array at processor boundaries
• MPI_PROC_NULL for boundaries in mpi_routines

[nadnaps]

• Matrix inversion for 4x4 matrices with a cleaner routine
• normals not updated before calculating bending forces
• update pressure at the end of substep loop
• check strain rate definition and allocation in mlsStruc

[nadnaps]

• In mlsibmtc - update add forces from cartesian version
• allreduce loop is better over particles than over vertices
• el_cx can be zero in MLS STRUC

[nadnaps]

• improve / remove hard-coded ray-tracing

[chongshenng]

spectra calculations for codes with periodic directions