EAMxx: clean up now-unneeded, then-needed barriers in shoc energy integrals

[mahf708]

Long story short: I was talking to @bartgol about something completely unrelated and I found the barriers like ones found in shoc_energy_integrals_impl.hpp to be quite surprising (then I noticed the note starting around

E3SM/components/eamxx/src/physics/shoc/impl/shoc_energy_integrals_impl.hpp

Line 29 in 958ad23

// The team_barriers protect what we think is unexpected behavior in

). After briefly discussing with @ambrad, these barrier may no longer be needed.

xref:

• issue and fixed in kokkos: Team parallel_reduce may need an impl-side team_barrier between adjacent reductions. kokkos/kokkos#5475
• for testing, Crusher: Nondeterministic seg faults in SCORPIO's flush_output_buffer's call to PnetCDF's ncmpi_wait_all scream#2080 (comment)
• scream pr: SHOC: Add some team_barriers that resolve a problem but don't make sense. scream#1921
• another scream pr: SHOC: Use another team_barrier based on view_reduction pattern. scream#1938

Proposed work/cleanup:

• remove the barriers
• conduct repeated testing on frontier/pm-gpu before merging