Memoization of rule arguments

[chavacava]

In the current implementation of rules, the configuration of rules is read and checked each time the rule is applied to a file.

For example in

revive/rule/banned-characters.go

Lines 17 to 33 in b331445

	func (r *BannedCharsRule) Apply(file *lint.File, arguments lint.Arguments) []lint.Failure {
	var failures []lint.Failure
	checkNumberOfArguments(1, arguments, bannedCharsRuleName)
	bannedCharList := r.getBannedCharsList(arguments)
	onFailure := func(failure lint.Failure) {
	failures = append(failures, failure)
	}
	w := lintBannedCharsRule{
	bannedChars: bannedCharList,
	onFailure: onFailure,
	}
	ast.Walk(w, file.AST)
	return failures
	}

lines 20 and 21 will be executed for each file with the very same arguments argument, thus the function calls will return always the same result.

This PR modifies the way argument of rules are handled by memoizing them and avoid checking and analyzing them for each file.

[mgechev]

Maybe == 0 for consistency?

[chavacava]

done

[mgechev]

Maybe == 0 for consistency?

[chavacava]

done

[mgechev]

LGTM. Left minor comments. Both > 1 and == 0 sound good. To me == 0 is a bit better, more deterministic.