Stop reserving 30 CPUs for a single-threaded lm

[rcannood]

Describe your changes

lm asks for highcpu (30 cores) but runs on one. The per-gene loop is

preds <- pbapply::pblapply(
  seq_len(ncol(X_mod2)),
  function(i) { ... }
)

with no cl argument, so it's serial. And the one thing that looks like it was meant to enable parallelism:

n_cores <- parallel::detectCores(all.tests = FALSE, logical = TRUE)

is computed at the top of the script and never referenced again.

Dropped the dead line and moved the label to lowcpu. Kept highmem/hightime -- the memory and runtime asks are real, it's only the CPU reservation that was fictional.

Happy to go the other way and pass cl = n_cores instead if we'd rather have the speedup; that's a bigger change though, since pblapply with a fork cluster would copy dr_train/X_mod2 per worker.

Part of a series of PRs coming out of a pre-run review of the benchmark.

Checklist before requesting a review

• I have performed a self-review of my code

• Check the correct box. Does this PR contain:

  • Breaking changes
  • New functionality
  • Major changes
  • Minor changes
  • Bug fixes

• Proposed changes are described in the CHANGELOG.md

• CI Tests succeed and look good!