[fix] modular setup macro padding

[luffykai]

Currently, if the modulus has 40 bytes, the blocksize is 8 and final number of limbs is still 40.
But we can't really do 40, as the adapter is either 32 or 48. Change the macro to match that logic

[github-actions]

group	app.proof_time_ms	app.cycles	app.cells_used	leaf.proof_time_ms	leaf.cycles	leaf.cells_used
verify_fibair	(-63 [-2.7%]) 2,233	513,774	19,386,023	-	-	-
fibonacci_program	(+35 [+0.6%]) 6,060	1,500,137	51,487,838	-	-	-
regex_program	(-34 [-0.2%]) 17,829	4,190,904	165,010,909	-	-	-
ecrecover_program	(+18 [+0.7%]) 2,613	285,401	15,075,033	-	-	-

Commit: f0502f0

Benchmark Workflow

[luffykai]

i am not sure if i understand this correctly, but since the limbs is either 32 or 48, i just make the block 32 or 16 for them (like the memory read block). lmk if this doesn't work

[Golovanov399]

Seems ok for now, but maybe at some point we need to do something to the adapters or at least to how many limbs we are using exactly in the operations

[jonathanpwang]

Seems ok for now, but maybe at some point we need to do something to the adapters or at least to how many limbs we are using exactly in the operations

Yes will revisit. @manh9203 can help track this issue.