Migrate cluster timestamp surfaces to Jiff

[danbi2990]

Background

Cluster timestamp handling spans multiple related surfaces:

• types::Cluster
• tables::Cluster
• private ClusterDbSchema

These are easy to confuse because the public DTO and stored table value are separate surfaces with similar timestamp fields.

Prerequisites

• Default bincode table value baseline has landed.
• Any cluster-specific public/stored fixture coverage has landed.
• The table-value implementation umbrella is open and linked.
• Old-byte fixture coverage for the exact tables::Cluster stored value exists before production type changes are made, or this issue adds that fixture coverage first.
• Public/API fixture or compile coverage for types::Cluster exists before production type changes are made, or this issue adds that coverage first.

Scope

• Migrate types::Cluster.last_modification_time.
• Migrate tables::Cluster.last_modification_time.
• Migrate private ClusterDbSchema.last_modification_time conversion/compile surface.
• Preserve stored bytes for tables::Cluster where possible.
• Preserve public API semantics for types::Cluster.

Acceptance criteria

• Public DTO and stored table value roles are reviewed separately.
• Old tables::Cluster stored bytes remain readable, or data migration is explicitly implemented.
• Public/API impact of types::Cluster is visible in tests or compile checks.
• Cluster query/conversion behavior still passes existing tests.
• Any remaining chrono usage in the touched cluster surfaces is explicitly justified.

Out of scope

• column_stats, even though it also uses NaiveDateTime.
• Final chrono dependency cleanup.

Notes

Related discussion: #733

[octoaide]

Hi — I can help move this migration forward but I need a couple quick confirmations before taking action.

Summary of the current situation I see:

• This issue is scoped to migrate last_modification_time across types::Cluster, tables::Cluster, and the private ClusterDbSchema conversion surface while preserving stored bytes where possible and keeping public DTO semantics intact.
• An assignee is set (sophie-cluml). I won't start coding unless you (or another maintainer) explicitly delegate or unassign the issue.
• Some prerequisites are mentioned in the issue. Add baseline tests for default bincode table values with chrono fields #746 (bincode baseline) is closed, but it’s not clear whether the other prerequisites (cluster-specific fixtures, old-byte fixtures for tables::Cluster, and the table-value umbrella PR/link) are in place.

Questions / requests

1. Can the current assignee (or a maintainer) confirm the status of the prerequisites listed in the issue? Specifically:

   • Are the cluster-specific public/stored fixtures already present? If so, please point to the test files/PR.
   • Is old-byte fixture coverage for the existing tables::Cluster stored value available? If not, do you want me to add it as part of this work?
   • Is the “table-value implementation umbrella” PR/issue open and linked? If yes, share the link so I can reference its approach.

2. Ownership/delegation: do you (sophie-cluml) want to keep this task, or would you like me to implement the migration? If you want me to implement, please either unassign yourself or explicitly say “please proceed” so I can set assignee_delegated = true and start coding.

Proposed minimal plan if implementation is approved

• Add/ensure old-byte fixtures for tables::Cluster so stored bytes remain readable (or add an explicit migration path if preservation is not feasible).
• Migrate tables::Cluster serialization to use Jiff while keeping the serialized layout compatible, or implement a data-migration when needed.
• Migrate types::Cluster API fields to Jiff, keeping public semantics the same and adding/adjusting compile/tests that show the impact.
• Update ClusterDbSchema conversion code and run compile/test coverage to ensure conversions work.
• Add tests: old-byte deserialization tests, public DTO compile/tests, and cluster query/conversion tests.

If you want, I can propose these as sub-issues (self-contained work items with tests included) for review — please react with a 👍 to this comment to approve that, or reply with any changes you prefer.

If you delegate implementation to me, I’ll:

• Create a single PR (or small set of PRs if the umbrella requires it) containing code+tests+fixtures, preserving stored bytes where possible, and request review from @sophie-cluml and @sehkone.

Thanks — please let me know the prerequisite status and whether I should proceed or hold for the assignee.