Optimize String#downcase and String#upcase for single byte optimizabl…

[asterite]

…e case

This is a very simple change that leads to a pretty good performance boost.

Here's the benchmark:

require "benchmark"

string = "hEllO WoRlD!." * 10

Benchmark.ips do |x|
  x.report("downcase") do
    string.downcase
  end
  x.report("upcase") do
    string.upcase
  end
end

Before:

downcase   7.73M (129.38ns) (± 1.92%)  144B/op        fastest
  upcase   7.68M (130.15ns) (± 2.50%)  144B/op   1.01× slower

After:

downcase  42.20M ( 23.69ns) (± 3.39%)  144B/op        fastest
  upcase  41.84M ( 23.90ns) (± 3.83%)  144B/op   1.01× slower

[straight-shoota]

Char#downcase/#upcase essentially does pretty much the same thing, just with a bit indirection through some method calls.

I suppose the performance gain comes from skipping the options.none? || options.ascii? test for every single byte, because the entire loop already is already guarded by that condition.

[asterite]

This interpreter spec is failing quite often:

    1) HTTP::Server handles Expect: 100-continue correctly when body is read
  
         server failed to start within 00:00:05 (Exception)

I guess the interpreter is slower and it takes more time than that? 🤔