[F] Implement libsoda ode solver

[kylebaron]

Summary

Replace ODEPACK ode solver with lsoda solver written in C++

https://github.com/dilawar/libsoda

FIrst step is to just get the simulation going with the new solver. I'd like to get it implemented alongside ODEPACK so we can still easily benchmark; but if it does well, tempted to just leave the FORTRAN code behind.

See pull request #519

• The odepack fortran code has been removed
• The only available solver is the C++ lsoda
• All of the same options can be passed through
  • hmax
  • hmin
  • atol and rtol (both implemented as scalar values for now)
  • maxsteps
• I refactored part of the LSODA object
  • Changed printf statements (and other message printing) to REprintf
  • Moved certain data items to be members of LSODA
  • Moved certain members to public for ease of access

The switch and code were discussed briefly with @chousand-metrum and @dpastoor

The switch was necessary due to repeated compiler warnings coming from the old ODEPACK code as CRAN moves to newer fortran compilers.

Tests

• tests/testthat/test-iv-po.R
  • CP from oral model is identical to closed form result
  • CP from iv model is identical to closed form result
• inst/validation/test-validation.R
  • non-stiff problem
  • stiff problem

[kylebaron]

The c++ solver worked really well. Deciding to make the switch.

This also removes odepack authors as package contributors and adds the author of the C++ lsoda code

/***

• Created: 2018-08-14

• Author: Dilawar Singh dilawars@ncbs.res.in

• Organization: NCBS Bangalore

• License: MIT License */

See . dc19112c2956498c172e6b4881040872384ffa81

[dpastoor]

LGTM