Refactor Time methods

[straight-shoota]

This PR refactors Time.year_month_day_year and Time.absolute_days to simplify the code and make it easier to understand.

The changes are mostly just formal, but there are a few semantic changes such as storing the value of an expression in a local variable for re-use.

[RX14]

Looks good, sorry to have to ask but do you have any performance benchmarks?

[straight-shoota]

year_month_day_day_year is about 25% faster than before.

I actually didn't expect much change in absolute_days but for some reason it's about 10% slower. But the measurements vary a lot, it's hard to nail it down, but that seems to be the trend.
When I change the operations back exactly how it was before, it somehow measures about 10% improvement.
This is all really weird and inconclusive. Can someone verify that?

This is my benchmark, it's running through the entire date range, so maybe a bit overkill...

require "benchmark"

struct Time
  Benchmark.ips do |bm|
    bm.report "new impl" do
      (1..9999).each do |year|
        ordinal = 0
        (1..12).each do |month|
          (1..days_in_month(year, month)).each do |day|
            ordinal += 1
            Time.absolute_days(year, month, day)
          end
        end
      end
    end
    bm.report "old impl" do
      (1..9999).each do |year|
        ordinal = 0
        (1..12).each do |month|
          (1..days_in_month(year, month)).each do |day|
            ordinal += 1
            Time.absolute_days_old(year, month, day)
          end
        end
      end
    end
  end
end

[RX14]

@straight-shoota 10% is fine