Add Censored distribution

[sethaxen]

This PR fixes #1468 for constructing a right-, left-, or interval-censored form of a distribution.

[codecov-commenter]

Codecov Report

Merging #1470 (f632f10) into master (cbbc1a5) will increase coverage by 0.54%.
The diff coverage is 99.62%.

@@            Coverage Diff             @@
##           master    #1470      +/-   ##
==========================================
+ Coverage   84.26%   84.80%   +0.54%     
==========================================
  Files         124      126       +2     
  Lines        7511     7778     +267     
==========================================
+ Hits         6329     6596     +267     
  Misses       1182     1182              

Impacted Files	Coverage Δ
src/truncate.jl	87.62% <ø> (ø)
src/censored.jl	99.61% <99.61%> (ø)
src/common.jl	79.59% <100.00%> (ø)
src/show.jl	97.67% <100.00%> (ø)
src/truncated/discrete_uniform.jl	100.00% <100.00%> (ø)
src/univariate/continuous/cauchy.jl	84.44% <0.00%> (+2.22%)	⬆️

---

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 cbbc1a5...f632f10. Read the comment docs.

[devmotion]

Thanks, I left some initial comments 🙂

[mschauer]

We could add the patch version increment for a release too.

[devmotion]

Did you build the documentation locally and checked that it looks fine?

[sethaxen]

Did you build the documentation locally and checked that it looks fine?

Yes, all looks good.

All changes have been incorporated and comments resolved, except for #1470 (comment).

With recent changes, I replaced some special-casing in the expectations with a utility function xexpy (similar in spirit to xlogy). This ensures everything works if the user does provide infinite bounds (we could strictly disallow this, but it could happen programmatically, and as mentioned somewhere above, it's in principle possible to have non-zero probability mass at infinity, e.g. Dirac(Inf)), hence providing an Inf bound is actually not the same thing as providing a nothing bound. This is now tested.

[sethaxen]

This should be updated to use the new truncation syntax before merging.

[sethaxen]

Okay, this should be up-to-date with master again.

[sethaxen]

@devmotion is anything else needed before merging this?

[devmotion]

Looks good, thanks for another great PR @sethaxen! I only have a minor suggestion regarding @check_args. I'll approve nevertheless, I'm fine with merging after it is addressed 🙂

[sethaxen]

@devmotion okay then, should be good to merge!