Improve index file header checkpoint logic in x/blockdb

[DracoLi]

Context and scope

Currently, we persist the index file header whenever the written block height is a multiple of CheckpointInterval. This approach is not ideal, as blocks can be written out of order during bootstrapping, potentially leading to concurrent calls to persist the index header. Instead, we can checkpoint after every x written blocks to smooth out the writes.

See #4027 (comment)