Fix issue with recompiling stan code

[WardBrian]

Summary:

I have backported the changes from stan-dev/stanc3#1074 into a 2.26 branch here. This PR changes the built stancjs to this version, which resolves an issue where subsequent calls to stancjs did not clean up prior inputs correctly.

This closes #982

Intended Effect:

Subsequent calls to stancjs should work even if earlier calls failed.

How to Verify:

Code provided by @andrjohns. Both calls fail on the current 2.26 version, only the first fails in this branch.

bad_mod <- "parameters { real no_semicolon } model {}"
good_mod <- "parameters { real semicolon; } model {}"

rstan::stanc(model_code = bad_mod)
rstan::stanc(model_code = good_mod)

Side Effects:

None

Documentation:

N/A

Reviewer Suggestions:

Copyright and Licensing

Please list the copyright holder for the work you are submitting (this will be you or your assignee, such as a university or company):
Simons Foundation

By submitting this pull request, the copyright holder is agreeing to license the submitted work under the following licenses:

• Code: GPLv3 (http://opensource.org/licenses/GPL-3.0)
• Documentation: CC-BY 4.0 (https://creativecommons.org/licenses/by/4.0/)

[WardBrian]

Note that it might be worthwhile adding the code in the PR text as a test, but I am not very familiar with the R unit testing standards

[hsbadr]

Thanks @WardBrian! I'll release new packages for testing and will work on revamping the unit tests in the experimental branch.