So with the help of Mathworks support, I was able to resolve the "ARM CMSIS SIN COS" error.
Running the following code from the command prompt fixed the issue:
RTW.TargetRegistry.getInstance('reset');
After this, for those that are interested, I had to do some interesting configuration edits on the STM side of things to have my 2014a setup working in 2019a matlab. First, I followed the instructions here and replaced the STM32/script/customRoutineProcess.tlc with the one in this forum post.
It also appears that in the change from 2018b to 2019a the "codeFormat" field is somehow not referenced correctly. This is referenced in the STM32/script/stm32_make_rtw_hook.m file. I figured this wasn't really a big deal, so I just commented line 202 out by adding a % sign. Essentially, it was this:
% Called after code generation is complete, and just prior to kicking
% off make process (assuming code generation only is not selected.) All
% arguments are valid at this stage.
fprintf('\n### Code Format : %s\n',buildOpts.codeFormat);
And I changed it to this:
% Called after code generation is complete, and just prior to kicking
% off make process (assuming code generation only is not selected.) All
% arguments are valid at this stage.
% fprintf('\n### Code Format : %s\n',buildOpts.codeFormat);