Refactor Interval{Float64}^x

[lbenet]

This PR uses only Float64 arithmetic to implement a method for ^(a::Interval{Float}, n::Integer), i.e., avoids using big53.

[coveralls]

Coverage decreased (-0.1%) to 92.181% when pulling 833158d on refactor_power into b6f3521 on master.

[codecov-io]

Codecov Report

Merging #47 into master will decrease coverage by 0.12%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master      #47      +/-   ##
==========================================
- Coverage   92.31%   92.19%   -0.13%     
==========================================
  Files          21       21              
  Lines         924      999      +75     
==========================================
+ Hits          853      921      +68     
- Misses         71       78       +7

Impacted Files	Coverage Δ
src/intervals/arithmetic.jl	98.75% <ø> (-0.02%)	⬇️
src/intervals/intervals.jl	96% <100%> (-4%)	⬇️
src/intervals/functions.jl	92.77% <100%> (-2.98%)	⬇️
src/intervals/rounding.jl	76.92% <0%> (+5.12%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update b6f3521...e12392e. Read the comment docs.

[dpsanders]

The problem is that the rounded versions of ^ still use BigFloat internally.

There is some kind of power available in CRlibm that I didn't wrap, but it is known that correctly rounding the power function is very difficult.

That is why the only solution I see is not to demand tight correct rounding.

[lbenet]

The problem is that the rounded versions of ^ still use BigFloat internally.

This PR replaces the version that uses big53 in ^(a::Interval{Float64}, n::Integer) by operations purely with Float64, which is the case considered by pow. The tests pass, so I would say it is a question of benchmarking against the previous version and pow. This version provides tight bounds as far as the ITF1788 tests are concerned.

At the moment I am playing with ``^(a::Interval{Float64}, n::Float64)` to see how far can I get.

[coveralls]

Coverage increased (+0.08%) to 92.395% when pulling 53a97ea on refactor_power into b6f3521 on master.

[lbenet]

This is ready for review; test pass locally!

[coveralls]

Coverage decreased (-0.1%) to 92.192% when pulling e12392e on refactor_power into b6f3521 on master.

[dpsanders]

I was referring to these lines, where ^(x, y, RoundDown) is defined in terms of BigFloats.

[lbenet]

You are totally right; I forgot the changes in rounding.jl...

[lbenet]

Closing this, to think this over...