Improve accuracy of sinpi and cospi

[OlivierHnt]

This PR is essentially a revival of the PR #421 which got forgotten unfortunately.
Closes #381; closes #412.

This PR also refactors the code in trigonometric.jl; the code is about 200 lines shorter.

[codecov-commenter]

Codecov Report

Attention: 7 lines in your changes are missing coverage. Please review.

Comparison is base (b702da4) 82.27% compared to head (b386b06) 84.76%.
Report is 1 commits behind head on master.

Files	Patch %	Lines
src/intervals/arithmetic/trigonometric.jl	97.01%	6 Missing ⚠️
src/intervals/rounding.jl	93.33%	1 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #600      +/-   ##
==========================================
+ Coverage   82.27%   84.76%   +2.49%     
==========================================
  Files          23       23              
  Lines        2138     2035     -103     
==========================================
- Hits         1759     1725      -34     
+ Misses        379      310      -69     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[OlivierHnt]

So it seems that the stack overflow comes from the fact that sinpi, and cospi are not properly defined for 32 bit system in CRlibm.jl.

Specifically, https://github.com/JuliaIntervals/CRlibm.jl/blob/d212ba9154ade995bb6c330c51092c685a8ec432/src/CRlibm.jl#L152-L155 do not include the functions sinpi, cospi, tanpi and atanpi.

I may be missing something here, but why were these functions omitted? At first glance, I would have defined function_names and MPFR_function_names to be the same.

cc @dpsanders

[OlivierHnt]

I fixed the StackOverflow issue for cospi and sinpi in 32 bit systems; I guess I was mislead into thinking that cospi and sinpi were defined in MPFR.