Fix % Normed{UInt32} on ARM and Improve NaN % FixedPoint

[kimikage]

Fixes #134 (#134 (comment))

julia> 1.0 % N0f32
0.4999999999N0f32 # before (on ARM)
1.0N0f32 # after

This also adds tests for NaN/Inf. This clarifies the behavior regarding NaN and Inf, which had not been explicitly defined.

This reduces the environment-/optimization-dependent instability of the results for NaN.

  Expression: NaN % N === NaN32 % N === zero(N)
   Evaluated: 65537.0N16f16 === 65537.0N16f16 === 0.0N16f16 # I could not reproduce this on my local machine.

(cf. https://github.com/JuliaMath/FixedPointNumbers.jl/runs/1029971789?check_suite_focus=true#step:5:26)
However, the rem is still affected by the undefined behavior of unsafe_trunc.

[codecov]

Codecov Report

Merging #223 into master will increase coverage by 0.50%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##           master     #223      +/-   ##
==========================================
+ Coverage   91.29%   91.79%   +0.50%     
==========================================
  Files           6        6              
  Lines         563      573      +10     
==========================================
+ Hits          514      526      +12     
+ Misses         49       47       -2     

Impacted Files	Coverage Δ
src/fixed.jl	95.48% <100.00%> (+0.21%)	⬆️
src/normed.jl	93.26% <100.00%> (+1.11%)	⬆️
src/utilities.jl	96.29% <100.00%> (+0.29%)	⬆️

---

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 9ca0d9d...da5f533. Read the comment docs.

[kimikage]

This is not the main topic of this PR, but the speed performance of % Fixed{Int8} was improved.

julia> x_f32 = rand(Float32, 1000, 1000);

julia> @btime $x_f32 .% Q0f7;
  8.012 ms (2 allocations: 976.70 KiB)   # v0.8.4
  762.600 μs (2 allocations: 976.70 KiB) # PR #220 
  125.699 μs (2 allocations: 976.70 KiB) # this PR

Of course this makes it easier to overflow as a trade-off, but still has enough range for practical purposes.

[kimikage]

This change is somewhat ugly, but I will merge this first as it is needed to implement the "new" division.