Consider being more selective in save() in prepare_distributed in run_future_lapply

[kendonB]

prepare_distributed contains the following call to save which is quite a bottleneck:

save(
    list = ls(globalenv(), all.names = TRUE),
    envir = globalenv(),
    file = globalenv_file(config$cache_path)
  )

Is it possible to refine this so that it only saves the items that are needed to build the targets to be made? This also seems unnecessary for many types of futures and seems to be replicating a function of the future package itself, which is really good at sending objects around different processes.

[wlandau]

We mostly need save() for Makefile parallelism. I think we can get away with just saving the imports. We can probably also suppress it altogether for "future" parallelism.

[wlandau]

I would prefer to handle this one after #440, specifically #440 (comment).

[wlandau]

f247d27 excludes targets in the plan from being saved. That is the best I can do safely.