Retire `Performance/Casecmp` cop

[parkerfinch]

This follows up on the ideas generated in #100.

It is confusing to have this cop enabled. Even when the autocorrection is marked as unsafe, the cop still directs people to use casecmp(...).zero?, which is incorrect in non-ascii use cases.

This issue pairs with rubocop/rubocop#9753, where I'm hoping we can create a Linting cop to suggest using casecmp?.

[mlegner]

There is another issue with the Performance/Casecmp cop: For str1.downcase == str2.downcase it suggests str1.casecmp(str2).zero?. But then Style/NumericPredicate jumps in to suggest str1.casecmp(str2) == 0 instead.

[Volosh1n]

Another issue with this cop:
str.#{selector} == ('string')
str.#{selector} != ('string')

Both lines above return the same offence message.

Checkout lines:

rubocop-performance/spec/rubocop/cop/performance/casecmp_spec.rb

Line 19 in 219bf57

^^^^^{selector}^^^^^^^^^^^^^^ Use `str.casecmp('string').zero?` instead of `str.#{selector} == ('string')`.

rubocop-performance/spec/rubocop/cop/performance/casecmp_spec.rb

Line 41 in 219bf57

^^^^^{selector}^^^^^^^^^^^^^^ Use `str.casecmp('string').zero?` instead of `str.#{selector} != ('string')`.