Is your feature request related to a problem? Please describe.
I have an (old) compiler/architecture where sizeof(int) is 16.
Trice relies on 32-bit integers when shifting IDs and the like. With explicit sizes, that is easily provided.
Trice also has some random magic numbers that are 32-bit. Restricting the values to use 16 bit should do the job as well.

Describe the solution you'd like
Use casts to fast-32-bit integers (mostly, uint_fast32_t) before shifting values, where 32 bit width are expected.
Use long integer literals (suffix 'l') where (at least) 32 bit width are expected.

Describe alternatives you've considered
A simple cast to uint32_t works, but might be slower on 64-bit architectures.
A #define to choose magic number width seems like overengineering.

Proposal
I will provide an appropriate pull request soon.