Add procedures to read and verify data from the advice stack

[hackaugusto]

We often have to write code that does the following:

• initialize the stack with the hasher state
• read N words from the advice stack to memory, and run a permute after each read
• extract the digest
• validate the data against a known value

I think steps 1 through 3 should have a procedure, and step the another one for steps 1 through 4.

[hackaugusto]

Well, I just realized we have some of this functionality already:

https://github.com/0xPolygonMiden/miden-vm/blob/a20cb5f87e8bb21bbad6e1d884a5a7950ef9a52b/stdlib/asm/mem.masm#L157-L164

But we never use it because that is less efficient, because the counter goes to the stack, and we usually write code with a hardcoded constants. I guess we have a few alternatives:

• Write multiple procedures, e.g. pipe_to_memory_2 / pipe_to_memory_4 / pipe_to_memory_6, where the number in the procedure is the number of words read from the advice stack to memory
• Extend the assembly syntax to allow for constant arguments to procedures, something like:

export.pipe_to_memory_even(count)
  repead.@count
    adv_pipe hperm
  end
end

Where the @count would be replaced by the assembler, similar to a preprocessor.

[hackaugusto]

The idea of using constant seems to have an issue of its own: https://github.com/0xPolygonMiden/miden-vm/issues/297

[bobbinth]

• Extend the assembly syntax to allow for constant arguments to procedures, something like:

export.pipe_to_memory_even(count)
  repead.@count
    adv_pipe hperm
  end
end

This is a bit tricky because for every value of @count, we'd have a different MAST root. This means that we won't be able to compile this function to MAST without knowing the calling context. The way the assembler is structured now, I believe something like that won't be possible, and also it won't be possible once we transition to MAST-based libraries. But overall, I agree, this would be a nice way to avoid some code duplication.