Add new Style/StringChars cop

## Summary

This PR adds new `Style/StringChars` cop. The cop checks for uses of
`String#split` with empty string or regexp literal argument.

```ruby
# bad
string.split(//)
string.split('')

# good
string.chars
```

Since Ruby 2.0, the behavior of `String#chars` and `String#split(//)` is the same.

### Ruby 1.9 and lower

```console
% ruby -ve "p 'foo'.split(//)"
ruby 1.9.3p551 (2014-11-13 revision 48407) [x86_64-darwin13.0.2]
["f", "o", "o"]
% ruby -ve "p 'foo'.chars"
ruby 1.9.3p551 (2014-11-13 revision 48407) [x86_64-darwin13.0.2]
#<Enumerator: "foo":chars>
```

### Ruby 2.0 and higher

```console
% ruby -ve "p 'foo'.split(//)"
ruby 2.0.0p648 (2015-12-16 revision 53162) [x86_64-darwin13.0.2]
["f", "o", "o"]
% ruby -ve "p 'foo'.chars"
ruby 2.0.0p648 (2015-12-16 revision 53162) [x86_64-darwin13.0.2]
["f", "o", "o"]
```

Writing `String#chars` would pretty express the intent of the code.

## Additional Information

This cop is intended to use `String#chars` method that express readable code.
And it has a different purpose than `Performance/RedundantSplitRegexpArgument` cop.
https://docs.rubocop.org/rubocop-performance/cops_performance.html#performanceredundantsplitregexpargument

Therefore, it will be added as a new `Style` cop.

changelog/new_add_new_style_string_chars_cop.md

@@ -0,0 +1 @@
+* [#9615](https://github.com/rubocop/rubocop/pull/9615): Add new `Style/StringChars` cop. ([@koic][])

config/default.yml

@@ -4493,6 +4493,12 @@ Style/StderrPuts:
   Enabled: true
   VersionAdded: '0.51'
 
+Style/StringChars:
+  Description: 'Checks for uses of `String#split` with empty string or regexp literal argument.'
+  Enabled: pending
+  Safe: false
+  VersionAdded: '<<next>>'
+
 Style/StringConcatenation:
   Description: 'Checks for places where string concatenation can be replaced with string interpolation.'
   StyleGuide: '#string-interpolation'

lib/rubocop.rb

@@ -582,6 +582,7 @@
 require_relative 'rubocop/cop/style/special_global_vars'
 require_relative 'rubocop/cop/style/stabby_lambda_parentheses'
 require_relative 'rubocop/cop/style/stderr_puts'
+require_relative 'rubocop/cop/style/string_chars'
 require_relative 'rubocop/cop/style/string_concatenation'
 require_relative 'rubocop/cop/style/string_hash_keys'
 require_relative 'rubocop/cop/style/string_literals'

lib/rubocop/cop/mixin/preferred_delimiters.rb

@@ -15,7 +15,7 @@ def initialize(type, config, preferred_delimiters)
       end
 
       def delimiters
-        preferred_delimiters[type].split('')
+        preferred_delimiters[type].chars
       end
 
       private

lib/rubocop/cop/style/command_literal.rb

@@ -165,7 +165,7 @@ def backtick_literal?(node)
         end
 
         def preferred_delimiter
-          (command_delimiter || default_delimiter).split('')
+          (command_delimiter || default_delimiter).chars
         end
 
         def command_delimiter

lib/rubocop/cop/style/regexp_literal.rb

@@ -150,7 +150,7 @@ def slash_literal?(node)
 
         def preferred_delimiters
           config.for_cop('Style/PercentLiteralDelimiters') \
-            ['PreferredDelimiters']['%r'].split('')
+            ['PreferredDelimiters']['%r'].chars
         end
 
         def correct_delimiters(node, corrector)

lib/rubocop/cop/style/string_chars.rb

@@ -0,0 +1,38 @@
+# frozen_string_literal: true
+
+module RuboCop
+  module Cop
+    module Style
+      # Checks for uses of `String#split` with empty string or regexp literal argument.
+      #
+      # This cop is marked as unsafe. But probably it's quite unlikely that some other class would
+      # define a `split` method that takes exactly the same arguments.
+      #
+      # @example
+      #   # bad
+      #   string.split(//)
+      #   string.split('')
+      #
+      #   # good
+      #   string.chars
+      #
+      class StringChars < Base
+        extend AutoCorrector
+
+        MSG = 'Use `chars` instead of `%<current>s`.'
+        RESTRICT_ON_SEND = %i[split].freeze
+        BAD_ARGUMENTS = %w[// '' ""].freeze
+
+        def on_send(node)
+          return unless node.arguments.one? && BAD_ARGUMENTS.include?(node.first_argument.source)
+
+          range = node.loc.selector.begin.join(node.loc.end)
+
+          add_offense(range, message: format(MSG, current: range.source)) do |corrector|
+            corrector.replace(range, 'chars')
+          end
+        end
+      end
+    end
+  end
+end

spec/rubocop/cop/style/string_chars_spec.rb

@@ -0,0 +1,54 @@
+# frozen_string_literal: true
+
+RSpec.describe RuboCop::Cop::Style::StringChars, :config do
+  it 'registers and corrects an offense when using `split(//)`' do
+    expect_offense(<<~RUBY)
+      string.split(//)
+             ^^^^^^^^^ Use `chars` instead of `split(//)`.
+    RUBY
+
+    expect_correction(<<~RUBY)
+      string.chars
+    RUBY
+  end
+
+  it "registers and corrects an offense when using `split('')`" do
+    expect_offense(<<~RUBY)
+      string.split('')
+             ^^^^^^^^^ Use `chars` instead of `split('')`.
+    RUBY
+
+    expect_correction(<<~RUBY)
+      string.chars
+    RUBY
+  end
+
+  it 'registers and corrects an offense when using `split("")`' do
+    expect_offense(<<~RUBY)
+      string.split("")
+             ^^^^^^^^^ Use `chars` instead of `split("")`.
+    RUBY
+
+    expect_correction(<<~RUBY)
+      string.chars
+    RUBY
+  end
+
+  it 'does not register an offense when using `chars`' do
+    expect_no_offenses(<<~RUBY)
+      string.chars
+    RUBY
+  end
+
+  it 'does not register an offense when using `split(/ /)`' do
+    expect_no_offenses(<<~RUBY)
+      string.split(/ /)
+    RUBY
+  end
+
+  it 'does not register an offense when using `split`' do
+    expect_no_offenses(<<~RUBY)
+      string.split
+    RUBY
+  end
+end