Fix checksum calculation following BIP-128 standard

[oren-z0]

Non-Ascii characters should not be converted for
checksum calculation.
This will give consistent hash to BIP-128 and its
Javascript code example.

Timelock-Recovery Plans that contained only ascii
characters are not affected.

[SomberNight]

Which field could contain non-ascii characters? I guess "wallet_name"?

This will give consistent hash to BIP-128 and its Javascript code example.

Are you saying the current python code does not produce the same hash as the example at the end of the BIP?
Why is that? Where is the non-ASCII character in that example?

A unit test would be nice.

[oren-z0]

Yes, "wallet_name" could contain non-ascii character.
The BIP also allows an optional field for a description and optional labels for the destinations, which are not implemented in Electrum, but could be implemented in the future (and contain non-ascii characters).

Yes, if using ensure_ascii=True does give a different result than Javascript's JSON.stringify().
With ensure_ascii=False the results are the same (tested using Python 3.13 and Node.JS 24.1).