datalake/translation: fix data partition to coordinator mapping

[bharathv]

An inherent assumption in coordinator implementation is that all topic/table related operations are performed by the same coordinator for correctness (so we have one active writer per table active at most times). This is particularly relevant in filesystem catalog implementation as it relies on the filesystem for concurrency and multiple interleaved writers can leave it in a questionable state.

This expectation is violated in our coordinator hash function that factors in topic and partition as input to the hash function resulting in ntps from same topic hashing to different coordinator partitions.

Backports Required

• [ ] none - not a bug fix
• [ ] none - this is a backport
• [x] none - issue does not exist in previous branches
• [ ] none - papercut/not impactful enough to backport
• [ ] v24.2.x
• [ ] v24.1.x
• [ ] v23.3.x

Release Notes

• none

  the below tests from https://buildkite.com/redpanda/redpanda/builds/57716#01930391-82d9-4800-92d2-1cef560a06b2 have failed and will be retried

  datalake_cloud_rpunit

  the below tests from https://buildkite.com/redpanda/redpanda/builds/57716#01930391-82db-47e1-89de-3b70dc092809 have failed and will be retried

  catalog_schema_manager_rpunit

[vbotbuildovich]

ducktape was retried in https://buildkite.com/redpanda/redpanda/builds/57716#019303e9-971d-4e3c-89c2-277b66d76b41 ducktape was retried in https://buildkite.com/redpanda/redpanda/builds/57716#019303e9-9720-43a7-ae3a-52af32bd2b13