fix rws padding logic and error handling

[hero78119]

Description

Fixed rws padding logic and padding error handling.

Issue Link

#1507

Type of change

• Bug fix (non-breaking change which fixes an issue)

Contents

• more documentation on padding logic
• fix max_rws calculation
• de-duplicated startop in end_block.tx by skip second one.
• panic instead of confused error log in padding calculation function.

[davidnevadoc]

Thanks for taking care of this! The comments and new variables make the calculations around the number of rows much easier to follow :)
LGTM 👍

[davidnevadoc]

I have a small question here:
Is there a valid scenario where we enter this branch? In other words, is it ok to have self.rows.len() == self.n_rows in some case?
I believe this is the situation where the log error was being triggered but the circuits were still correct.

[hero78119]

y here is exactly the case self.rows.len() == self.n_rows Because state_circuit receive RwMap which is padding ready, also we set self.n_rows == FixedCParams.max_rws that's why here is strictly equal.

Besides, the result of padding_length is not used for padding here. It's for override, because override idx is counting after padding rows.

I will rename padding_length to first_non_padding_index instead

[hero78119]

Fixed to have better naming in commit 6d5fced

[ed255]

LGTM! The comments in the code are very useful to explain how the extra rows are calculated :)