Unreachable code in ML-DSA

[abhinav-thales]

The code to calculate pending elements(<256) of a matrix cell after a 5 block 'squeeze' in ML-DSA has unreachable code as the buflen is a multiple of 3 due to STREAM128_BLOCKBYTES(168)

Raised a issue here

[SWilson4]

Thanks for the report, @abhinav-thales. If this was detected via a static analysis tool or another automated test, it would be great to have it in our CI.