Improve accuracy of delta_h for large hue differences

[kimikage]

This is a follow-up to PR #501.
This avoids the loss of significance when |Δh| > 90 deg.
This also adds delta_c to improve accuracy of Lab{Float32} chroma differences.

julia> a, b = Lab(60, -36.75124f0, -4.8861504f0), Lab(80, 4.557669f0, -2.6522517f0);

julia> Colors.delta_h(a, b)
-26.459148f0 # master
-26.459143f0 # this PR

julia> Colors.delta_h(Lab{BigFloat}(a), Lab{BigFloat}(b))
-26.45914388977270124791437895470628506022253442451544986697371328558951044881593

julia> chroma(a) - chroma(b)
31.801413f0

julia> Colors.delta_c(a, b)
31.801414f0

julia> chroma(Lab{BigFloat}(a)) - chroma(Lab{BigFloat}(b))
31.80141452063303790359376637963574569550424247180931525255208635027325242378461

[codecov]

Codecov Report

Merging #503 (c4dafa0) into master (f38f27c) will increase coverage by 0.04%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #503      +/-   ##
==========================================
+ Coverage   93.60%   93.64%   +0.04%     
==========================================
  Files           9        9              
  Lines        1111     1118       +7     
==========================================
+ Hits         1040     1047       +7     
  Misses         71       71              

Impacted Files	Coverage Δ
src/utilities.jl	98.68% <100.00%> (+0.04%)	⬆️

---

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 f38f27c...c4dafa0. Read the comment docs.

[kimikage]

A: 2 * muladd(c1, c2, -dp)
B: muladd(dc, -dc, muladd(da, da, db^2))

When comparing A and B, A is generally more accurate because it only subtracts once. However, even A can lose digits, in which case B can be more accurate. 😕

Edit: I retuned it in PR #494. 😅
Edit2: I also fixed the typo in the comment about the talyor series. 😭