New cop: Use String#include? instead of String#match? with /literal-only-pattern/

[pocke]

Is your feature request related to a problem? Please describe.

String#include? is faster than String#match? with a simple regexp.

https://gist.github.com/pocke/eeffeaa58ddcc35f862f6e1aca593b21
I wrote a few benchmarks, and I got the result.

The difference is small in simple cases (str1 and str2).
But there are clear differences if the string is long.

Describe the solution you'd like

Add a cop to detect String#match? with simple regexp.
And I guess the cop also can detect String#match and String#=~ if it is used as a condition.

Describe alternatives you've considered

nothing

Additional context

We already have Performance/StartWith and EndWIth. We can refer the cops to help to implement it.

[Drenmi]

And I guess the cop also can detect String#match and String#=~ if it is used as a condition.

It's very hard to tell if people are using the global variables stored by String#match, but perhaps can include a configuration option for this?

[pocke]

It's very hard to tell if people are using the global variables stored by String#match, but perhaps can include a configuration option for this?

Good point! I forgot the global variables.
Performance/RegexpMatch can be aware of the global variables, so the include cop will be able to support them also by using Performance/RegexpMatch code.

But I guess we can just omit supporting match and =~ from include cop.
RegexpMatch cop corrects match to match if it is possible, so the include cop can detect match + simple regexp usage after running RegexpMatch cop even if the include cop does not support match and =~ directly.

[bbatsov]

Agreed. Generally the use of the globals has been on the decline, so I guess we can also advise people not to do it. I guess the only thing the globals have for them is that they are more compact than the alternative.