Create MLJ compliant doc strings

[ablaom]

• Some items flagged "TODO", waiting on clarification at Help with documentation review #136.

@tlienart

Note to self: We should probably add MLJTestInterface tests before tagging next release.

[codecov-commenter]

Codecov Report

Merging #138 (7c9e435) into dev (30f7a30) will increase coverage by 0.65%.
The diff coverage is 0.00%.

❗ Current head 7c9e435 differs from pull request most recent head bd0a20f. Consider uploading reports for the commit bd0a20f to get more accurate results

📣 This organization is not using Codecov’s GitHub App Integration. We recommend you install it so Codecov can continue to function properly for your repositories. Learn more

@@            Coverage Diff             @@
##              dev     #138      +/-   ##
==========================================
+ Coverage   94.62%   95.28%   +0.65%     
==========================================
  Files          20       21       +1     
  Lines         875      869       -6     
==========================================
  Hits          828      828              
+ Misses         47       41       -6     

Impacted Files	Coverage Δ
src/mlj/classifiers.jl	100.00% <ø> (+50.00%)	⬆️
src/mlj/doc_tools.jl	0.00% <0.00%> (ø)
src/mlj/interface.jl	85.45% <0.00%> (+4.42%)	⬆️
src/mlj/regressors.jl	12.50% <ø> (+5.83%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[tlienart]

nit: Its

[ablaom]

Good catch. Addressed.

[ablaom]

Thanks for all the help, @tlienart. This PR is now ready for review. Please pay particular attention to the solver field documentation.

One remaining point of confusion for me is this: I'm interpreting the existing docs and what your have helped with as "ISTA and FISTA are disallowed if gamma = 0" but I think you only mean to say these are the only options when gamma > 0 (because we need a non-smooth solver in that case) - and not that it can't still be used when gamma = 0 as well (although better options exist in that case). So, in this PR, I've said (F)ISTA solvers are allowed for any model.

[tlienart]

One remaining point of confusion for me is this: I'm interpreting the existing docs and what your have helped with as "ISTA and FISTA are disallowed if gamma = 0" but I think you only mean to say these are the only options when gamma > 0 (because we need a non-smooth solver in that case) - and not that it can't still be used when gamma = 0 as well (although better options exist in that case). So, in this PR, I've said (F)ISTA solvers are allowed for any model.

This is correct, with gamma=0 the prox operation will be the identity and it will be a kind of gradient descent which should generally be ok (I've not tested but I don't see a reason why it would fail; I'll open an issue to test this explicitly)

[tlienart]

LGTM with minor changes; feel free to merge after that

[tlienart]

I imagine you might want to do a patch release with this, let me know if you need my help in doing so (I think you have full admin access to this repo though). Thanks for going through this @ablaom !

[ablaom]

Just going to check #140 to be sure.

[ablaom]

Before the patch we should address #141