Increase default pool_size to 4

[npwalker]

Solves #1038. pool_size was added as an optional setting
a while ago and people have been using it successfully
in the wild. This commit increases the default to 4
which seems like a generally useful improvement.

Please add all notable changes to the "Unreleased" section of the CHANGELOG.

[mwaggett]

I'm curious how 4 was decided on.

[npwalker]

I chose 4 as two users have indicated they use 5 successfully. The linked issue #1038 shows some time breakdowns for one user's testing.

I'd be willing to compromise to 3 or 2 if we are worried about 4 but given modern computing 4 threads to download and extract modules seems relatively safe. I'd be surprised if it performed worse than 1 in any production scenario.

[adrienthebo]

@npwalker There's a bug in one of the tests; we're copying references to a single object instead of having for objects. Try applying this patch:

diff --git i/spec/unit/action/puppetfile/install_spec.rb w/spec/unit/action/puppetfile/install_spec.rb
index 904983b..f77641d 100644
--- i/spec/unit/action/puppetfile/install_spec.rb
+++ w/spec/unit/action/puppetfile/install_spec.rb
@@ -19,7 +19,9 @@ describe R10K::Action::Puppetfile::Install do
 
   describe "installing modules" do
     let(:modules) do
-      Array.new(4, R10K::Module::Base.new('author/modname', "/some/nonexistent/path/modname", nil))
+      (1..4).map do |idx|
+        R10K::Module::Base.new("author/modname#{idx}", "/some/nonexistent/path/modname#{idx}", nil)
+      end
     end
 
     before do

[adrienthebo]

Regarding the level of concurrency, 4 is a reasonable value. I'm 👍 once the tests are green.

[reidmv]

👍 to upping the default.