RuboCop improvements

[AlexWayfer]

• Remove disabling Style/AccessModifierDeclarations RuboCop cop It's OK to place module_function above all module functions.
• Enable new cops from RuboCop
• Remove disabling Metrics/MethodLength RuboCop cop for specs Resolve offenses, improve specs a bit.
• Merge inherited Include and Exclude in RuboCop config
• Add required_ruby_version to gemspec Remove disabling Gemspec/RequiredRubyVersion RuboCop cop.
• Remove enabling Style/ArrayCoercion RuboCop cop Why? If it's disabled by default — this is better for community, I think.
• Remove enabling RuboCop cops, already enabled by default

[AlexWayfer]

As this PR is inspired by #11, @dblock can be interested in.

[michaelherold]

Merged via 7ee4886.

Thanks for the contribution, @AlexWayfer!

As a bit of friendly advice, I suggest asking before making sweeping style-related changes to projects. The way you rewrote a bunch of tests made it harder to track the history of the file. Change history is much more important than adhering to draconian style rules from an external tool. In particular, the BlockLength cop is :-1: on any test files that use RSpec because they are, inherently, long blocks due to how the tool works!

I appreciate the effort you put into the changes, so accepted most of them squashed into a single commit. I also took the liberty of bumping to the Ruby versions supported by the latest version of memory_profiler, which, now that I think about it, might have broken the Travis build. Next up, I'm going to look at the GitHub Actions branch, so I don't consider that a problem.

[AlexWayfer]

As a bit of friendly advice, I suggest asking before making sweeping style-related changes to projects.

It's a PR to open-source, you owe nothing to me, I understand risks before making changes. It's just a suggestion and can be discussion.

The way you rewrote a bunch of tests made it harder to track the history of the file. Change history is much more important than adhering to draconian style rules from an external tool.

1. I don't think it's a lot of changes, but OK, it's subjectively.
2. I can agree that change history is important, but I can't agree that it's very important. Especially I don't see dirt in 1–2 commits, there are possibilities to blame further.
3. It's not just "an external tool" for me and many others, it's a community standard, also flexible (configurable), so we can deal with it in desirable way.

In particular, the BlockLength cop is on any test files that use RSpec because they are, inherently, long blocks due to how the tool works!

Yep, I can agree, that's why I'm adding spec/**/* into BlockLengths Exclude in almost every project. It's OK for me: RuboCop doesn't force styles for any of external gems (maybe only with plugins).

I appreciate the effort you put into the changes, so accepted most of them squashed into a single commit.

Once again: I don't afraid to remake my PR, remove some commits, change them, etc. So, if something — don't hesitate to ask, please, at the end — it's my proposal.