Support softmax for row_vector, matrix types + code simplifications

[jachymb]

Summary

This refactors the softmax and log_softmax functions in the following ways:

• both vector and row_vector are now supported instead of just vector
• matrix type is also supported and the softmax is applied row-wise. (i.e. the return value is effectively a row_stochastic_matrix for modelling purposes)
• refactored the relevant code to utilize Eigen functionalities instead of some for loops and some other bits here and there, simplifying the code and improving clarity. I am honestly not a C++ guru, but I believe these shouldn't have any negative impact on performance.

Tests

• Added test for the new types
• Added a test for the edge case when some of the values is -inf

Side Effects

None to my knowledge.

Release notes

softmax and log_softmax now support row_vector and matrix (applied row-wise)

Checklist

• Copyright holder: Me, jachymb@gmail.com
  - Code: BSD 3-clause (https://opensource.org/licenses/BSD-3-Clause)
  - Documentation: CC-BY 4.0 (https://creativecommons.org/licenses/by/4.0/)

• the basic tests are passing

  • unit tests pass (to run, use: ./runTests.py test/unit)
  • header checks pass, (make test-headers)
  • dependencies checks pass, (make test-math-dependencies)
  • docs build, (make doxygen)
  • code passes the built in C++ standards checks (make cpplint)

• the code is written in idiomatic C++ and changes are documented in the doxygen

• the new changes are tested

[jachymb]

I think it's some server error, not actual test failure.

[SteveBronder]

@serban-nicusor-toptal Can you take a look at https://jenkins.flatironinstitute.org/blue/organizations/jenkins/Stan%2FMath/detail/PR-3313/4/pipeline/ ?

I'm not sure what is going on with the nohup permissions there

EDIT: Fixed