refactor(merkle): prevent overlapping chunk ranges in incremental trie updates

[AshinGau]

Description

Problem

The incremental merkle update loop uses range.step_by(incremental_threshold) combined with inclusive range start_block..=chunk_to, which causes boundary blocks to be processed twice.
Before (log / overlapping):

2025-12-03T03:48:20.240579Z  DEBUG Processing chunk current=50000 target=90000 incremental_threshold=7000 chunk_range=50001..=57001
2025-12-03T03:48:20.270721Z  DEBUG Processing chunk current=50000 target=90000 incremental_threshold=7000 chunk_range=57001..=64001
2025-12-03T03:48:20.290994Z  DEBUG Processing chunk current=50000 target=90000 incremental_threshold=7000 chunk_range=64001..=71001
2025-12-03T03:48:20.331801Z  DEBUG Processing chunk current=50000 target=90000 incremental_threshold=7000 chunk_range=71001..=78001
...

Solution

Replace step_by iterator with a while loop that advances start_block to chunk_to + 1, ensuring non-overlapping ranges.
After (log / non-overlapping):

2025-12-03T03:51:54.325227Z  DEBUG Processing chunk current=90000 target=150000 incremental_threshold=7000 chunk_range=90001..=97001
2025-12-03T03:51:54.356900Z  DEBUG Processing chunk current=90000 target=150000 incremental_threshold=7000 chunk_range=97002..=104002
2025-12-03T03:51:54.384736Z  DEBUG Processing chunk current=90000 target=150000 incremental_threshold=7000 chunk_range=104003..=111003
2025-12-03T03:51:54.414019Z  DEBUG Processing chunk current=90000 target=150000 incremental_threshold=7000 chunk_range=111004..=118004
...

Note

The current merkle algorithm is idempotent, so correctness is unaffected. This fix eliminates redundant computation and ensures correctness if the implementation ever becomes non-idempotent.

[mediocregopher]

This looks right to me, will wait for @shekhirin