fix: save S3 listing in precompute step, avoid loading s3fs in build step

s3fs uses reticulate/Python which significantly increases R process memory.
When pak tries to fork R to install package dependencies, the enlarged
process can't be duplicated within the container memory limit.

Now the install-deps step saves the S3 file listing as s3_cache.rds,
and the build step reads it with readRDS() — no s3fs loading needed.

local/packages-to-build.R

@@ -83,6 +83,10 @@ s3_pkgs = s3fs::s3_dir_ls(
   recurse = TRUE
 )
 
+# Save the raw S3 file listing for the build step to use as s3_package_cache
+# This avoids loading s3fs/reticulate in the build container, saving memory for pak forks
+saveRDS(basename(s3_pkgs), "/mnt/cache/packages/s3_cache.rds")
+
 file_names <- basename(s3_pkgs)
 matches <- regexec("^([A-Za-z0-9.]+)_([0-9][^/]*)\\.tar\\.gz$", file_names)
 parts <- regmatches(file_names, matches)