Normal format for custom_insn_* macros we have

[Golovanov399]

This resolves INT-2667.

The following is done here:

• custom_insn_r and custom_insn_i macros (are now procedural and) now have a clearer signature, where all arguments must be named, and the register types also have to be explicitly specified,
• technically, these macros generate the code under the #[cfg(target_os = "zkvm")] block, which would help the developer to catch the wrong macro signature things instantly if they have the language server running, but in reality we also want to let uninit = ... under the same block anyway, so now this change is not really represented. Maybe we will want to make uninit an additional valid macro argument at some point, but this would overcomplicate things imo.

[jonathanpwang]

Overall this is a big improvement!
LGTM, but just the macro use of opcode = const #opcode to insert assembly is a little weird to me.
There is one const vs sym question.

Also, does anything in the book/docs need to be updated, or do we not even mention this macro in the book / documentation?

Heads up I will likely merge #1107 first, so there will be a small(hopefully) rebase necessary.

[github-actions]

group	app.proof_time_ms	app.cycles	app.cells_used	leaf.proof_time_ms	leaf.cycles	leaf.cells_used
verify_fibair	(-15 [-0.4%]) 3,822	746,050	30,008,524	-	-	-
fibonacci_program	(+48 [+0.8%]) 6,077	1,500,137	51,505,102	-	-	-
regex_program	(+683 [+3.7%]) 19,197	4,190,904	165,028,173	-	-	-
ecrecover_program	(+17 [+0.6%]) 2,645	285,401	15,092,297	-	-	-

Commit: 5ab6d6b

Benchmark Workflow

[Golovanov399]

Btw yes, I think we don't mention these macros anywhere in the docs