revert: remove stress-test seeding offset changes

[mmagician]

Summary

• Reverts the index offset separation (pub_offset/priv_offset) and high-entropy StdRng-based seed generation in the stress test seeding
• Returns to the original simpler block_num * num_accounts + account_index indexing and index.to_be_bytes() seed generation

These changes were introduced to avoid account ID prefix collisions, but the stress test passes without them (verified locally with seed-store --num-accounts 500 --public-accounts-percentage 50).

Test plan

• Stress test seed-store passes locally without these changes

🤖 Generated with Claude Code

[mmagician]

LGTM

[Mirko-von-Leipzig]

Maybe to note that we need this to work with millions of accounts as well. So collisions are maybe possible?

[mmagician]

Maybe to note that we need this to work with millions of accounts as well. So collisions are maybe possible?

Yes we might end up with some collisions w.h.p. for large account numbers. This PR only reverts the changes introduced (by confusion in the other PR to keep the diff there relevant).

Are you suggesting to keep the changes? If so, I'd maybe leave this for a follow-up, where we also adapt the stress test in CI?

[Mirko-von-Leipzig]

Maybe to note that we need this to work with millions of accounts as well. So collisions are maybe possible?

Yes we might end up with some collisions w.h.p. for large account numbers. This PR only reverts the changes introduced (by confusion in the other PR to keep the diff there relevant).

Are you suggesting to keep the changes? If so, I'd maybe leave this for a follow-up, where we also adapt the stress test in CI?

Ah, I wasn't aware of the context :) Was just saying that 500 accounts isn't sufficient to be sure :D Happy to merge this then.