Optimize validation rule hot paths

[swalkinshaw]

Extracted from #5578

These a few miscellaneous micro-optimizations to validation rules.

• ArgumentNamesAreUnique: avoid Hash.new block allocation on the common unique-args path; only build array on actual collision
• FieldsHaveAppropriateSelections: fast-path leaf+no-selections and non-leaf+selections cases, skipping the full validation method
• RequiredArgumentsArePresent: cache required arg names per definition to avoid re-iterating arguments for repeated field definitions
  - FragmentSpreadsArePossible: use Array#intersect? instead of none? { include? } for type overlap check

Note: I'm keeping all the FieldsWillMerge optimizations for their own PR.

[swalkinshaw]

Oh didn't realize Ruby 2.7 was still supported 😓 so intersect? isn't available. I'll drop that one

[rmosolgo]

Thanks for these improvements!

I'll have to keep intersect in mind for when we drop Ruby 2.7 support. I think there are other places it could be used, too, for example:

graphql-ruby/lib/graphql/execution/next/runner.rb

Line 383 in 6d1e5fa

p_types.any? { |t| c_p_types.include?(t) }

graphql-ruby/lib/graphql/analysis/query_complexity.rb

Line 163 in 6d1e5fa

query.types.possible_types(b).any? { |t| a_types.include?(t) }