Unify multi-condition filtering

[pirj]

Always use all? semantic when applying metadata conditions for filtering.
Previously all? was only used for hooks, and any? was used for everything else (module inclusions, shared example group inclusion, define derived metadata).

Before this change MyModule would be included to example groups defining any of foo: true or bar: true:

RSpec.configure do |c|
  c.include MyModule, :foo => true, :bar => true
end

After the change the same any? behaviour is achievable splitting into two statements:

RSpec.configure do |c|
  c.include MyModule, :foo => true
  c.include MyModule, :bar => true
end

MyModule will only be included in groups with both :foo => true AND :bar => true

RSpec.configure do |c|
  c.include MyModule, :foo => true, :bar => true
end

Fixes #1821

Depends on:

• Update ammeter (to a temporary repo) to support RSpec 4 rspec-rails#2468
• Work around ammeter' incompatibility with RSpec 4 rspec-rails#2469

[pirj]

The failure is due to https://github.com/alexrothenberg/ammeter/blob/master/lib/ammeter/rspec/generator/example.rb#L14 expecting any? semantic.
I'll send a PR.

[JonRowe]

Why? (not had enough ☕ yet 😂 )

[pirj]

This was the hardest part of the change.
Without this, exclusions.include_example?(ex) would just always return true, and its negation - false.
This is due to the change in the underlying logic in the case when there are no rules, that is coming from this:

[].all?(&:odd) # new version => true
[].any?(&:odd) # before => false

Without this check, rspec always finishes with 0 examples, 0 failures.

[pirj]

InclusionFilter has a similar check:

      def include_example?(example)
        @rules.empty? || super
      end

I thought of refactoring ExclusionRules to a similarly defined exclude_example?, but left it if the later refactoring. Especially since !exclude_example? is hard to wrap your head around quickly when called on an instance of ExclusionRules 😄

[JonRowe]

I'm having a think about wether include should take this into account, this just seems like a trap currently.

[pirj]

Do you think there could be several rules in the exclusion filter, and we shouldn't check against them with all??

[JonRowe]

I feel like this should be encapsulated somewhere, but for now I think this is fine

[pirj]

Green!

[pirj]

This is finally the last one change on the TODO list for 4.0, please take a look. :begger:

What's left is to add deprecation warnings for what we've changed/removed in 4.0 to 3.x.

I've made several attempts of removing WithKeywordsWhenNeeded with no luck removing any. I can keep bashing my head against it, or we can release 4.0 and deal with kwargs after. Up to you to decide, @JonRowe