Cleanup Variables API

[alyst]

Another round of cherry-picks from #193 that cleans up the variables API, as discussed in #199:

• vars()/nvars() methods to get the vector of SEM model variables/count of variables.
• observed_vars()/nobserved_vars() to get the vector of observed variables ordered as in observed data (nobserved_vars() replaces n_man() method)
• latent_vars()/nlatent_vars() to get the vector of latent variables, preserving their order in vars()
• get_data() renamed to samples(), nsamples() is the number of data points/samples/observations in the observed data (nsamples() replaces n_obs() method)
• for SemMissingData nmeasured_vars() is the count of observed variables with measurements (within the given data pattern)
• interface method definitions are moved from types.jl to the files, where the methods of a particular type are defined (new files are created to accommodate methods for the abstract classes, e.g. SemSpecification)
• these variable API methods should work not only for the SEM specification objects, like RAMMatrices or ParameterTable, but also for the types that refer to SEM specification (SemImply, AbstractSem etc)
• old methods using ambiguous terms get_colnames(), or get_n_nodes() removed

Per #199 discussion, an alternative to observed/latent terms would be manifest/latent. Also, samples could be replaced with observations (keeping both nobserved_vars() and nobservations() may lead to confusion), and measured/missing terms could be replaced with observed/missing. I really don't have a strong preference here, so if SEM stakeholders think manifest/latent + observed/missing is a better choice, I can update the PR.
Also, I think obs_cov/obs_mean have to be updated accordingly (observed_cov/observed_mean or manifest_cov/manifest_mean).

This PR should be really the last one that does not introduce improvements or new features, but it should help to make the improvements easier.

[codecov]

Codecov Report

Attention: Patch coverage is 91.27907% with 15 lines in your changes missing coverage. Please review.

Project coverage is 70.52%. Comparing base (b2012b0) to head (2f6e8b7).
Report is 1 commits behind head on devel.

Files	Patch %	Lines
src/frontend/specification/RAMMatrices.jl	71.42%	4 Missing ⚠️
src/observed/EM.jl	80.00%	4 Missing ⚠️
src/frontend/specification/StenoGraphs.jl	80.00%	2 Missing ⚠️
src/frontend/common.jl	80.00%	1 Missing ⚠️
src/frontend/fit/standard_errors/bootstrap.jl	0.00%	1 Missing ⚠️
src/frontend/fit/summary.jl	0.00%	1 Missing ⚠️
src/imply/RAM/generic.jl	93.33%	1 Missing ⚠️
src/imply/RAM/symbolic.jl	85.71%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##            devel     #207      +/-   ##
==========================================
+ Coverage   69.64%   70.52%   +0.88%     
==========================================
  Files          51       52       +1     
  Lines        2421     2446      +25     
==========================================
+ Hits         1686     1725      +39     
+ Misses        735      721      -14     

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

[alyst]

@Maximilian-Stefan-Ernst I have also cleaned up the existing unit tests and added the tests to cover the vars/params API calls

[Maximilian-Stefan-Ernst]

Thanks again for this great PR, and also for adding/restructuring tests. Sorry for taking so long to review. If the comments are resolved, this is ready to merge; I also agree with the terminology.

[alyst]

@Maximilian-Stefan-Ernst Thank you for the review. I think I've addressed your comments: params, vars and observed APIs are now exported, StenoGraph explicit dependency is removed from the tests, as well as the alternative check_acyclic() method.
Params suggestions look like false positives (again). I wonder if the new v3 version of julia-format action fixes that.

[Maximilian-Stefan-Ernst]

Thanks, I will update julia-format...