Condense seed dispersal update in elm_drv with a fates procedure

[glemieux]

The regular seed dispersal update code in elm_drv should be neatly wrapped up into fates code.

• move mpi call into a new elmfates_interface procedure
• refactor bc_in and bc_out assignment wrap_seed_... calls into elmfates_interfaceMod, instead of elm_driver
• create "dispersal" type to replace the elm_driver seed_ local variable
• Update seed dispersal code to accommodate multiple pfts fates#23
• move allocation of seed distribution arrays out of elm_driver into an initialization routine
  - [ ] Add global variable to set the distribution frequency (check against is_beg_curr_...)
• Add fates-side mode selection variable/logic to turn dispersal on or off

[glemieux]

Order of operations:

Init:

1. Allocate bc_in/out via allocate_bcin/bcout (called from elmfates init <- elm_inst_biogeochem)
2. Zero out bc_in(s)%seed_in and bc_out(s)%seed_out via zero_bcs (called from elmfates init <- elm_inst_biogeochem)

Dynamics:

1. Calculate and set bc_out%seed_out(pft) via call to SeedIn (called from ed_integrate_state_variables <- ed_ecosystem_dynamics <- dynamics_driv)
2. Send seed from bc_out%seed_out into an local output buffer for the current gridcell
3. Distribute seed information from current gridcell to all gridcells (wrapped mpi call in elm_drv out OMP thread area)
4. Accumulate incoming seed from global gridcell-sized array and pass back into bc_in%seed_in values via wrap_seed_dispersal

[glemieux]

Unless we are planning on sharing the seed information with the host land model, we don't really need this to be formatted as part of the bc_in/out data structure. That said, this setup will provide future-proofing if so desired. The wrapped mpi call does need to take place within elm_drv outside of the OMP threading pragmas, but we could still use separate fates type structure.

[glemieux]

I need to refresh myself on how the probability density is used in conjunction with the seed mass. This PD value for each neighboring grid cell in held in the neighborhood type and not used currently. Since its held in the neighborhood_type and indexed by gridcell, the calculation needs to happen in WrapSeedGlobal outside of the fates threaded area.

• Review application of PD functions

UPDATE: Yes, I need to update the incoming_global value based on the PD. See glemieux/fates#23 for further updates.

[glemieux]

Complete